研究者詳細 - 齊藤　剛史

論文 - 齊藤　剛史

分割表示 >> ／全件表示180 件中 1 - 180 件目

Efficient DNN Model for Word Lip-Reading 査読有り国際誌

Arakane T., Saitoh T.

Algorithms 16 ( 6 ) 2023年06月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper studies various deep learning models for word-level lip-reading technology, one of the tasks in the supervised learning of video classification. Several public datasets have been published in the lip-reading research field. However, few studies have investigated lip-reading techniques using multiple datasets. This paper evaluates deep learning models using four publicly available datasets, namely Lip Reading in the Wild (LRW), OuluVS, CUAVE, and Speech Scene by Smart Device (SSSD), which are representative datasets in this field. LRW is one of the large-scale public datasets and targets 500 English words released in 2016. Initially, the recognition accuracy of LRW was 66.1%, but many research groups have been working on it. The current the state of the art (SOTA) has achieved 94.1% by 3D-Conv + ResNet18 + {DC-TCN, MS-TCN, BGRU} + knowledge distillation + word boundary. Regarding the SOTA model, in this paper, we combine existing models such as ResNet, WideResNet, WideResNet, EfficientNet, MS-TCN, Transformer, ViT, and ViViT, and investigate the effective models for word lip-reading tasks using six deep learning models with modified feature extractors and classifiers. Through recognition experiments, we show that similar model structures of 3D-Conv + ResNet18 for feature extraction and MS-TCN model for inference are valid for four datasets with different scales.

DOI： 10.3390/a16060269

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85163730596&origin=inward

Detection of Script Reading Face in Diet Deliberation Video 査読有り

Yamaguchi T., Saitoh T., Zhang H.

Lecture Notes in Electrical Engineering 1322 LNEE 189 - 198 2025年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper presents a method for analyzing Japanese Diet deliberations video footage to determine if a speaker reads from a script while speaking. The study focuses on classifying script reading faces as a preprocessing step for future research. Four main classification methods are proposed: (1) using the success or failure of face detection, (2) utilizing head pose angle based on the PnP problem, (3) employing head pose angle based on a regression model, and (4) using an image recognition model. Twelve videos from a Video Retrieval System for Diet Deliberations were used, and 30-s speaking scenes were extracted to evaluate the proposed approaches. The results indicate that approach (4) achieved the highest classification accuracy.

DOI： 10.1007/978-981-96-1535-3_20

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85219209279&origin=inward

LGNMNet-RF: Micro-Expression Detection Using Motion History Images 査読有り国際誌

Teng M.K.K., Zhang H., Saitoh T.

Algorithms 17 ( 11 ) 2024年11月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

Micro-expressions are very brief, involuntary facial expressions that reveal hidden emotions, lasting less than a second, while macro-expressions are more prolonged facial expressions that align with a person’s conscious emotions, typically lasting several seconds. Micro-expressions are difficult to detect in lengthy videos because they have tiny amplitudes, short durations, and frequently coexist alongside macro-expressions. Nevertheless, micro- and macro-expression analysis has sparked interest in researchers. Existing methods use optical flow features to capture the temporal differences. However, these optical flow features are limited to two successive images only. To address this limitation, this paper proposes LGNMNet-RF, which integrates a Lite General Network with MagFace CNN and a Random Forest classifier to predict micro-expression intervals. Our approach leverages Motion History Images (MHI) to capture temporal patterns across multiple frames, offering a more comprehensive representation of facial dynamics than optical flow-based methods, which are restricted to two successive frames. The novelty of our approach lies in the combination of MHI with MagFace CNN, which improves the discriminative power of facial micro-expression detection, and the use of a Random Forest classifier to enhance interval prediction accuracy. The evaluation results show that this method outperforms baseline techniques, achieving micro-expression F1-scores of 0.3019 on CAS(ME)2 and 0.3604 on SAMM-LV. The results of our experiment indicate that MHI offers a viable alternative to optical flow-based methods for micro-expression detection.

DOI： 10.3390/a17110491

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85210317376&origin=inward

KuchiNavi: Lip-reading-based Navigation App 査読有り

Kanamaru T., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 13089 2024年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Lip-reading technology has the advantage that it can be used even in noisy environments and has been actively studied in recent years. In this paper, we develop a navigation application, "KuchiNavi," as a new application using lip-reading technology. The basic technology is word-level lip-reading technology, which utilizes an existing deep-learning model. However, we quantitatively evaluated lip-reading accuracy by selecting words for navigation, collecting utterance scenes independently, building an original dataset, and conducting recognition experiments. This paper, 101 Japanese words were selected, utterance scenes were collected from 15 people, and recognition experiments were conducted using the speaker-independent recognition task, the leave-one-person-out method. As a result, an average recognition rate of 88.2% was obtained. In addition, we developed an iOS app and conducted a demonstration in a car to confirm its effectiveness.

DOI： 10.1117/12.3021118

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85191656750&origin=inward

Joint Attention Detection Using First-Person Points-of-View Video 査読有り

Bise K., Saitoh T., Tsuchiya K., Sato H., Nakamura K., Abe T., Coffey F.

Proceedings - 2024 2nd International Conference on Computer Graphics and Image Processing, CGIP 2024 118 - 123 2024年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Both verbal and nonverbal communication is important to facilitate collaborative activities by multiple people. This study focused on eye gaze as a form of nonverbal communication. The proposed method consists of a local feature-matching method, an object detection method, and a combination of the two. A simulated scene was designed and filmed to evaluate the proposed method quantitatively. Experimental results show that the proposed method can achieve high accuracy when applied to local feature-matching. Furthermore, the effectiveness of the proposed method was confirmed by shooting a real scene of a medical procedure in a hospital.

DOI： 10.1109/CGIP62525.2024.00029

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85195118963&origin=inward

Efficient Construction of Face Image Dataset for Head Pose Estimation 査読有り国際誌

Furuya Y., Saitoh T., Tsuchiya K., Sato H., Nakamura K., Abe T., Coffey F.

Proceedings - 2024 2nd International Conference on Computer Graphics and Image Processing, CGIP 2024 175 - 180 2024年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

The purpose of this study is to estimate the head pose of a person in an operating room. Most publicly available datasets for head pose estimation (HPE) are images taken from the front, making it difficult to estimate the head pose of the back of the head. In addition, since they are not intended to healthcare professional's head pose, they do not include scenes in which a person is wearing a medical cap and a mask. Therefore, existing datasets do not match the purpose of our project and cannot correctly estimate head pose. Therefore, in this paper, we construct our original dataset. Since constructing a dataset is labor-intensive, we propose a method to construct a dataset in which data is distributed efficiently and uniformly using a turntable. A dataset consisting of approximately 20,000 images was constructed using the proposed method. We evaluated the accuracy of HPE using an existing deep learning model and obtained MAE of 5.67° for the pitch angle and 6.37° for the yaw angle. Furthermore, we compared the results with models trained on other datasets for an actual operating room scene and showed that our constructed dataset is useful.

DOI： 10.1109/CGIP62525.2024.00038

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85195128888&origin=inward

Construction of a New Database for Japanese Lip-reading 査読有り国際誌

Yamabe T., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 13089 2024年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

In recent years, lip-reading techniques have been actively researched for estimating speech content only from visual information without audio information. Large databases are available for English but not enough for other languages. Therefore, this paper constructs a new database for improving the accuracy of Japanese lip-reading. In previous research, we asked collaborators to record utterance scenes to build a database. This paper uses YouTube videos. We download a weather forecast video from the “Weathernews” YouTube channel. We constructed a database that can be used for lipreading by applying video and audio processing. Furthermore, we selected 50 Japanese words from our database and applied an existing deep-learning model. As a result, we obtained a word recognition rate of 66%. We have established a method for constructing a lip-reading database using YouTube, although there are still problems with the scale of the database and recognition accuracy.

DOI： 10.1117/12.3021119

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85191654714&origin=inward

Performance Improvement of Multi-Robot Data Transmission in Aggregated Robot Processing Architecture with Caches and QoS Balancing Optimization 査読有り国際誌

Jalil A., Kobayashi J., Saitoh T.

Robotics 12 ( 3 ) 2023年06月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

Robot Operating System 2 (ROS 2) is a robotic software that uses a set of Quality of Service (QoS) policies to manage the quality of robot data transmissions in a network, such as the RELIABLE and KEEP_LAST options. In ROS 2 node communication, the RELIABLE connection guarantees that all message data can be properly sent from the publisher to the subscriber. However, strict reliability is not guaranteed if the RELIABLE connection uses the KEEP_LAST option to transmit the robot data in the publish–subscribe communication. This study aims to analyze the efficiency of local cache, cache control, and QoS balancing optimization to improve ROS 2 node communication when using the RELIABLE and KEEP_LAST options to transmit multi-robot data in Aggregated Robot Processing (ARP) architecture. Our idea in local cache and cache control is to streamline the sensor data output before processing it when the sensor device produces the data with the same value in a row. Furthermore, QoS balancing optimization aims to balance the DEPTH and DEADLINE QoS configuration to determine the rates and buffer size in ROS 2 node communication. This study shows that combining local cache and QoS balancing optimization improves multi-robot data transmission and cooperation in ARP architecture.

DOI： 10.3390/robotics12030087

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85163380224&origin=inward

Optimization Algorithm for Balancing QoS Configuration in Aggregated Robot Processing Architecture 査読有り国際誌

Jalil Abdul, Kobayashi Jun, Saitoh Takeshi

人工生命とロボットに関する国際会議予稿集 ( 株式会社ALife Robotics ) 28 919 - 923 2023年02月

担当区分：最終著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Quality of Service (QoS) manages the data traffic to reduce packet loss, latency, and jitter in the network. This study aims to design an optimization algorithm to find the balance of QoS configuration to set the rates and buffer size while the robot data processes are communicated in the Aggregated Robot Processing (ARP) architecture. This study implements optimization to manage the DEPTH and DEADLINE QoS configuration in Robot Operating System 2 (ROS 2) node communication. Unbalancing DEPTH and DEADLINE configurations can affect the high latency time of message data transmission and packet loss in RELIABLE connections. The results of this study show that the optimization algorithm can determine the optimal value of DEPTH and DEADLINE by balancing the QoS configuration to improve the robot data transmission in the ARP architecture.

DOI： 10.5954/icarob.2023.gs4-4

Isolated single sound lip-reading using a frame-based camera and event-based camera 査読有り国際誌

Kanamaru T., Arakane T., Saitoh T.

Frontiers in Artificial Intelligence 5 2023年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

Unlike the conventional frame-based camera, the event-based camera detects changes in the brightness value for each pixel over time. This research work on lip-reading as a new application by the event-based camera. This paper proposes an event camera-based lip-reading for isolated single sound recognition. The proposed method consists of imaging from event data, face and facial feature points detection, and recognition using a Temporal Convolutional Network. Furthermore, this paper proposes a method that combines the two modalities of the frame-based camera and an event-based camera. In order to evaluate the proposed method, the utterance scenes of 15 Japanese consonants from 20 speakers were collected using an event-based camera and a video camera and constructed an original dataset. Several experiments were conducted by generating images at multiple frame rates from an event-based camera. As a result, the highest recognition accuracy was obtained in the image of the event-based camera at 60 fps. Moreover, it was confirmed that combining two modalities yields higher recognition accuracy than a single modality.

DOI： 10.3389/frai.2022.1070964

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146780678&origin=inward

Vision Transformer-Based Bark Image Recognition for Tree Identification 査読有り国際誌

Yamabe T., Saitoh T.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 13836 LNCS 511 - 522 2023年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Our group is studying tree species recognition using image processing technology. In the previous research, we proposed an image-based bark recognition using CNN. In this paper, we propose a method of recognizing bark image using Vision Transformer (ViT), which has attracted attention in the image recognition task in recent years. Four public datasets of NewBarkTex, TRUNK12, BarkNet1.0, and Bark-101, and a new dataset of 150 tree species originally collected, KyutechBark150, were used in the evaluation experiment. Several CNN models were used as comparison methods. As a result of the recognition experiment, the highest recognition accuracy of ViT was obtained in all the datasets. In addition, the trained model was visualized by t-SNE and attention map, and this paper shows that ViT is effective for bark image recognition.

DOI： 10.1007/978-3-031-25825-1_37

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85148044090&origin=inward

Optimization Algorithm for Balancing QoS Configuration in Aggregated Robot Processing Architecture 査読有り国際誌

Jalil A., Kobayashi J., Saitoh T.

Proceedings of International Conference on Artificial Life and Robotics 928 - 932 2023年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Quality of Service (QoS) manages the data traffic to reduce packet loss, latency, and jitter in the network. This study aims to design an optimization algorithm to find the balance of QoS configuration to set the rates and buffer size while the robot data processes are communicated in the Aggregated Robot Processing (ARP) architecture. This study implements optimization to manage the DEPTH and DEADLINE QoS configuration in Robot Operating System 2 (ROS 2) node communication. Unbalancing DEPTH and DEADLINE configurations can affect the high latency time of message data transmission and packet loss in RELIABLE connections. The results of this study show that the optimization algorithm can determine the optimal value of DEPTH and DEADLINE by balancing the QoS configuration to improve the robot data transmission in the ARP architecture.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85149458445&origin=inward

Conformer-Based Lip-Reading for Japanese Sentence 査読有り国際誌

Arakane T., Saitoh T., Chiba R., Morise M., Oda Y.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 13836 LNCS 474 - 485 2023年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Various applications of lip-reading technology are considered, and it is an important issue. In recent years, lip-reading research on sentences has attracted attention. However, most of the published datasets are English-talking scenes, and there are few datasets other than English. Therefore, in this research, we are researching Japanese sentence-level lip-reading. In this paper, we construct Japanese sentence utterance scene datasets ITA and ROHAN4600 and propose the Conformer-based lip-reading method. Recognition experiments were conducted using the Transformer model as a conventional method. As a result, it was confirmed that the Conformer model obtained high recognition accuracy both at the phoneme and the mora levels.

DOI： 10.1007/978-3-031-25825-1_34

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85148010751&origin=inward

QoS Balancing Optimization in Aggregated Robot Processing Architecture: Rate and Buffer 査読有り国際誌

Jalil Abdul, Kobayashi Jun, Saitoh Takeshi

Journal of Advances in Artificial Life Robotics ( 株式会社 ALife Robotics ) 3 ( 4 ) 209 - 213 2023年01月

担当区分：最終著者記述言語：英語掲載種別：研究論文（学術雑誌）

This study developed an optimization algorithm to balance the Quality of Service (QoS) rates and buffer size for robot data communication in the Aggregated Robot Processing (ARP) architecture. Robot Operating System 2 (ROS 2) is robotic software that uses a set of QoS policies to control the quality of data transmission in robotic networks, such as DEADLINE to determine the rates, and DEPTH to determine the buffer size. An unbalanced DEADLINE and DEPTH configuration in ROS 2 node communication can result in high packet latency and packet loss in RELIABLE connections. This happens when the DEADLINE sets the rates at high frequencies, and the DEPTH sets the buffers with small sizes. The results of this study show that the optimization algorithm developed in this study can determine the rate and buffer size through a balancing configuration for better quality robot data transmission in ARP architecture, influence the improvement of latency and reduce the packet loss.

DOI： 10.57417/jaalr.3.4_209

Can you read lips with a masked face? 査読有り国際誌

Arakane T., Kai C., Saitoh T.

Proceedings of MVA 2023 - 18th International Conference on Machine Vision and Applications 2023年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

We have been working on lip-reading that estimates the content of utterances using only visual information. When most people started wearing masks due to the coronavirus pandemic, several people asked us if using machine-based lip-reading even when wearing a mask was possible. Taking this as a research question, we worked on word-level lip-reading for the world's first masked face image. The utterance scene dataset when wearing a mask is not open to the public, so we developed our dataset, face detection for masked face images, facial landmarks detection, and lip-reading based on deep learning. We collected speech scenes of 20 people for 15 Japanese words and obtained a recognition accuracy of 88.3% as a result of the recognition experiment. This paper reports that lip-reading is possible for masked face images.

DOI： 10.23919/MVA57639.2023.10215925

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85170547632&origin=inward

Action request episodes in trauma team interactions in Japan and the UK - A multimodal analysis of joint actions in medical simulation 査読有り国際誌

Tsuchiya K., Coffey F., Nakamura K., Mackenzie A., Atkins S., Chałupnik M., Whitfield A., Sakai T., Timmons S., Abe T., Saitoh T., Taneichi A., Vernon M., Crundall D., Fuyuno M.

Journal of Pragmatics 194 101 - 118 2022年06月

記述言語：英語掲載種別：研究論文（学術雑誌）

Grounding is a fundamental human practice for cooperation and collaboration in a joint activity, when more than two people interact. Emergency care is one such interactive situation, and whether a trauma team can efficiently establish and increment their common ground at an appropriate timing during the complex and fluid activity of emergency medical treatment is key to maximise collective competence to best perform as a trauma team. This article investigates recurrent patterns in the grounding process between the trauma team leader and the members, comparing the practices between Japan and the UK, using an eye-tracking device. The embodied practice of grounding was multimodally described, applying both quantitative multimodal corpus analytic and qualitative interactional linguistic approaches. The analysis has shown that five grounding episodes reoccurred, most of which were more ego-centric and one of them ba-centric interactions, drawing on intersubjectivity and the theory of ba in Western and Eastern philosophy respectively.

DOI： 10.1016/j.pragma.2022.04.009

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85130373499&origin=inward

3DCNN-based Mouth Shape Recognition for Patient with Intractable Neurological Diseases 査読有り国際誌

Nakamura Y., Saitoh T., Itoh K.

Proceedings of SPIE - The International Society for Optical Engineering 12083 2022年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Patients with intractable neuropathy may use a mouth-shape character (MSC)-based communication as an alternative to communication when speech, writing, and PC are unavailable. MSC-based communication requires not only a speaker but also a supporter. However, it may not be easy to read the mouth shape due to the skill of the supporter. The communication support system that automatically recognizes the mouth shape of a speaker is expected. This study aims to develop the whole support system, and this paper works on mouth shape recognition. We introduce 3DCNN-based mouth shape recognition. As for the input data of CNN, not only the color image but also the flow image obtained by applying the optical flow is used, and two outputs of two CNN models are integrated. We collected speech scenes of eight patients with intractable neurological diseases and conducted recognition experiments. As a result, an average recognition rate of 77.1% was obtained. Excluding the two patients, one had difficulty recognizing the mouth shape even by a human due to little movement of the mouth, and the other had a problem shooting, and the average recognition rate of 86.6% was obtained. We demonstrated the effectiveness of the proposed method.

DOI： 10.1117/12.2623642

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85125590419&origin=inward

Replacing Speaker-independent Recognition Task with Speaker-dependent Task for Lip-reading Using First Order Motion Model 査読有り

Kodama M., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 12083 2022年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

There is a tendency to deal with a speaker-independent recognition task in the lip-reading field by collecting speech scenes from many speakers. The data collection task is time-consuming. This paper proposes a method to solve this problem. According to a driving video, First Order Motion Model (FOMM) is a deep generative model that generates a video sequence from a source image. Our idea is to apply FOMM to all speech scenes in the dataset to generate the speech scenes recording from one speaker. We propose a preprocessing method to replace the speaker-independent recognition task with the speaker-dependent recognition task by applying FOMM. We applied the proposed method to two publicly available databases: OuluVS and CUAVE, and confirmed that the recognition accuracy was improved by applying the proposed method to both databases.

DOI： 10.1117/12.2623640

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85125595947&origin=inward

Japanese sentence dataset for lip- reading 査読有り国際誌

Shirakata T., Saitoh T.

Proceedings of MVA 2021 - 17th International Conference on Machine Vision Applications 2021年07月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This research is about lip-reading for Japanese sentences. Research on English sentences is actively pursued due to the extensive datasets. However, a sufficient dataset for Japanese sentences has not been released. Therefore, this paper builds a Japanese sentence dataset. A Transformer model is used for the recognition task. Three recognition target levels: phoneme, mora, and vowel, are set, and recognition experiments show that they can be recognized.

DOI： 10.23919/MVA51890.2021.9511353

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85113921382&origin=inward

Multimodal Corpus Analysis of English Presentations for EFL teaching: Speech Contents, Speakers' Behavior and Gaze Points of Audience 査読有り国際誌

Miharu Fuyuno, Takeshi Saitoh

12th International Conference on Corpus Linguistics (CILC2021) 22 - 23 2021年04月

Night Time Calling System by Eye Movement using Wearable Camera 国際誌

Kazuki Sakamoto, Takeshi Saitoh, Kazuyuki Itoh

8th International Joint Symposium on Applied Engineering and Sciences (SAES2020) 2020年12月

Leader Identification Using Multimodal Information in Multi-party Conversations 査読有り

Shiota T., Honda K., Shimada K., Saitoh T.

2020 International Conference on Asian Language Processing, IALP 2020 7 - 12 2020年12月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

It is one of the important tasks to predict a participant's role in a multi-party conversation. Many previous studies utilized only verbal or non-verbal features to construct models for the role recognition task. In this paper, we propose a model that combines verbal and non-verbal features for leader identification. We add non-verbal features and construct our prediction model with utterance, pose, facial, and prosodic features. In our experiments, we compare our model with a baseline model that is based on only utterance features. The results show the effectiveness of our multimodal approach. In addition, we improve the performance of the baseline model to add some new utterance features.

DOI： 10.1109/IALP51396.2020.9310465

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85099875287&origin=inward

眼球運動を利用した呼び出しシステムの開発招待有り査読有り

齊藤剛史，坂本一樹，伊藤和幸，白井誠

地域ケアリング ( 北隆館 ) 22 ( 12 ) 2020年10月

Lip reading using facial expression features 査読有り国際誌

Tatsuya Shirakata, Takeshi Saitoh

4th International Conference on Imaging, Vision & Pattern Recognition (IVPR2020) 2020年08月

神経難病患者向けの口形の動きを利用したコミュニケーション支援に関する研究招待有り査読有り

齊藤剛史，中村祐哉，伊藤和幸，白井誠

地域ケアリング ( 北隆館 ) 22 ( 10 ) 2020年08月

Lip reading using facial expression features 査読有り国際誌

Tatsuya Shirakata, Takeshi Saitoh

International Journal of Computer Vision & Signal Processing 10 ( 1 ) 9 - 15 2020年07月

Gaze-Point Analysis of EFL Learners while Watching English Presentations: Toward Effective Teaching 査読有り国際誌

Miharu Fuyuno, Takeshi Saitoh, Yuko Yamashita, Daisuke Yokomori

International Journal for Educational Media and Technology 14 ( 1 ) 17 - 28 2020年06月

A team leader’s gaze before and after making requests in emergency care simulation: a case study with eye-tracking glasses 査読有り国際誌

Kyota Nakamura, Takuma Sakai, Takeru Abe, Takeshi Saitoh, Frank Coffe, Andrew MacKenzie, Akira Taneichi, Keiko Tsuchiya

BMJ Simulation and Technology Enhanced Learning 6 ( 6 ) 369 - 370 2020年03月

Automatic tree species identification from natural bark image 査読有り国際誌

Ido J., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 11373 2020年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2019 SPIE. In this research, we are studying on image-based identification of trees species that we can see everywhere. In our previous study, we showed that convolutional neural network (CNN) can recognize tree species by using a region of interest (ROI) image of bark. However, the bark region is manually extracted from a natural bark image. This paper solves this problem using semantic segmentation, and proposes an automatic tree species identification from natural bark image. The proposed method was evaluated with the bark image dataset collected independently. We confirmed the effectiveness of the proposed method.

DOI： 10.1117/12.2557187

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079816202&origin=inward

Development of Night Time Calling System by Eye Movement Using Wearable Camera 査読有り

Sakamoto K., Saitoh T., Itoh K.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 12426 LNCS 345 - 357 2020年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper proposes a night time call system using a wearable camera for patients. The proposed system consists of a wearable camera, computer, relay controller, and nurse call. The user wears the wearable camera. All captured eye images are fed to the convolutional neural network to detect the pupil center. When the detected pupil center exceeds a preset threshold value, the computer sends the signal to operate the nurse call via each relay controller. Two experiments were conducted to evaluate the proposed system: verification of the accuracy of pupil center detection and quantitative evaluation of call success. In the former experiment, we collected 2,800 eye images from seven people and conducted a pupil center detection experiment on several training conditions. As a result, an average error of 1.17 pixels was obtained. In the latter experiment, a call experiment was conducted on five healthy people. The experiment time for each subject was about five minutes. The subject experimented while lying on the bed. Twelve audio stimuli were given in one experiment, after getting the stimuli, the subject moved his eye. The correct call in response to the audio stimulus was considered successful, and the precision, recall, and F-measure were calculated. As a result, we obtained the precision, recall, and F-measure of 0.83, 1.00, and 0.91, respectively. These experimental results show the effectiveness of the proposed system.

DOI： 10.1007/978-3-030-60149-2_27

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85092207616&origin=inward

救急医療での非言語行為解析のためのチームリーダーの視線分析

土屋慶子, 中村京太, 安部猛, 齊藤剛史

電子情報通信学会技術研究報告 119(415) 33 - 36 2020年01月

Lip Reading using Facial Expression Features 査読有り国際誌

齊藤剛史

International Journal of Computer Vision and Signal Processing 1 9 - 15 2020年01月

Lip Reading Experiments for Multiple Databases using Conventional Method 査読有り

Shirakata T., Saitoh T.

2019 58th Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2019 409 - 414 2019年09月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2019 The Society of Instrument and Control Engineers - SICE. Several databases have been published in the field of lipreading. Each database has a wide variety of utterance content, number of speakers, shooting angles, and so on. However, each paper uses one or two databases. It is desirable to discuss using many databases. This paper uses four publicly available databases: CUAVE, OuluVS, CENSREC-1-AV, and SSSD. In this paper, not the latest deep learning-based method but the standard recognition method by hidden Markov model (HMM) which mainly used conventionally is applied, and analyzes trends in recognition accuracy. Based-on recognition experiments, it was found that the recognition accuracy was correlated with the number of frames.

DOI： 10.23919/SICE.2019.8859932

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85073875139&origin=inward

LiP25w: Word-level Lip Reading Web Application for Smart Device 査読有り国際誌

Takeshi Saitoh, Michiko Kubokawa

5th International Conference on Auditory-Visual Speech Processing (AVSP2019) 2019年08月

Automatic coding of a team leader’s eyegaze in emergency care simulation: developing a multimodal corpus with the sceneAnalysis GUI 査読有り国際誌

Tsuchiya, K, Saitoh, T, Nakamura, K, Sakai, T, Abe, T, Taneichi, A

The 10th International Corpus Linguistics Conference (CL2019) 2019年07月

A leader’s request and rapport in emergency care simulation: a multimodal corpus analysis 査読有り

Keiko Tsuchiya, Akira Taneichi, Kyota Nakamura, Takuma Sakai, Takeru Abe, Takeshi Saitoh

日本ヘルスコミュニケーション学会雑誌 ( 日本ヘルスコミュニケーション学会 ) 10 ( 1 ) 36 - 41 2019年06月

Dynamic Body Circumference Measurement Technique for a More Realistic Virtual Fitting Room Experience 査読有り

Tomi B., Sunar M., Mohamed F., Saitoh T., Mokhtar M., Luis S.

2018 IEEE Conference on e-Learning, e-Management and e-Services, IC3e 2018 56 - 60 2019年01月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. A body measurement technique for virtual apparel by using angle-based algorithm is presented. The existing virtual apparel or virtual fitting room system only focus on normal-sized human body. This paper proposes a technique that will allow the system to measure accurately and dynamically, to be used by users with various body sizes, using a depth motion sensor. This paper will also discusses about the data from the depth motion sensor, and how the data is processed using the proposed angle-based algorithm and ellipse perimeter equation. Experiments from this study have shown that the system can measure various human body size and the calculated measurement using the technique is almost the same as the actual measurement of the user's body. Potential application of this technique and future plan are also discussed.

DOI： 10.1109/IC3e.2018.8632636

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85062847742&origin=inward

CNN-based tree species identification from bark image 査読有り国際誌

Ido J., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 11069 2019年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only. This paper proposes a convolutional neural network (CNN)-based tree species identification method from bark image. The proposed method uses the well-known CNN model. The difficulty of our problem is to use a special tree image in which a colorful tag is stick on the bark. The tag is irrelevant to the species. In order to recognize with CNN, it is necessary to extract a region (ROI) excluding the tag. Thus, this paper proposes a ROI extraction method. Extracted ROI is fed to CNN. We evaluated the proposed method with six tree species. We carried out the evaluation experiment with various conditions, and found an optimal condition for our problem.

DOI： 10.1117/12.2524213

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85065800899&origin=inward

Character input system based on gaze input using wearable camera 査読有り国際誌

Chinsatit W., Kondo N., Saitoh T.

Proceedings of SPIE - The International Society for Optical Engineering 11069 2019年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only. This research presents the character input system by using wearable camera based gaze estimation system (GES) for disorders. The GES is the system that can estimate the user's gaze direction. The proposed system uses the inside-out camera to get the user's gaze image. This system uses the gaze image to classify the character and create the word. Proposed system consists of two processes: Gazes estimation process and character recognition process. This research uses the Convolutional Neural Network (CNN) method to archive the good accuracy. By using the proposed method to train the model, the system was confirmed through evaluation experiments. The experimental results show the good accuracy of the proposed system compared with the previous research.

DOI： 10.1117/12.2524411

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85065789895&origin=inward

英語プレゼンテーション動画視聴時の日本人英語学習者の注視点解析

冬野美晴, 山下友子, 齊藤剛史

本教育メディア学会年次大会プロシーディング 26 127 - 128 2019年01月

A leader’s request and rapport in emergency care simulation: a multimodal corpus analysis. 査読有り

土屋慶子, 安部猛, 齊藤剛史

The Journal of the Japanese Association of Health Communication. 10(1) 36 - 41 2019年01月

SSSD: Speech Scene database by Smart Device for Visual Speech Recognition 査読有り

Saitoh T., Kubokawa M.

Proceedings - International Conference on Pattern Recognition 2018-August 3228 - 3232 2018年11月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. Speech scenes of conventional databases available for lip reading or visual speech recognition (VSR), were record with a video camera fixed on a tripod in a well-maintained environment. On the other hand, VSR is expected to be used in smart devices such as smartphone and tablet as an interface. Therefore, collecting the speech scenes recorded with these devices is an important task for practical use of VSR. In this paper, we collect Japanese word utterance scenes taken with smart device, and build a new publicly available database named speech scene database by smart device (SSSD) for VSR. Moreover, we apply the existing method to our database and show baseline recognition accuracy.

DOI： 10.1109/ICPR.2018.8545664

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85059762033&origin=inward

Effectiveness of Data Augmentation for CNN-Based Pupil Center Point Detection 査読有り

Kan N., Kondo N., Chinsatit W., Saitoh T.

2018 57th Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2018 41 - 46 2018年10月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 The Society of Instrument and Control Enginners - SICE. Convolutional neural network (CNN) is one of the appearance-based methods, and this method has a potential for pupil center point detection problem. However, CNN requires a very large amount of training data. To overcome this problem, data augmentation (DA) is useful solution. In this paper, we investigate the effects of three DA approaches; affine transformation-based approach, synthetic image-based approach, and generative adversarial network (GAN)-based approach, and evaluate the effectiveness of DA for pupil center point detection task using an off-the-shelf CNN model. We set eleven conditions of the training data, and evaluated the detection accuracy. The result showed that higher detection accuracy was obtained for eye image by applying affine transformation. In our pupil center point detection task, it is found that affine transformation-based approach of DA is effective.

DOI： 10.23919/SICE.2018.8492593

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85056711036&origin=inward

Excretion Prediction Using Nursing Record System Log Data 査読有り

Saitoh T., Yamada I., Yoshioka Y.

2018 57th Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2018 1618 - 1623 2018年10月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 The Society of Instrument and Control Enginners - SICE. Excretion assistance that helps excretion is one of the most important care innursing care, however, there are many troubles concerning excretion at nursing care sites. This paper proposes a novel method to extract features from log data of nursing record system and to discriminate excretion or non-excretion for improving quality of life of care recipient and reducing work burden of caregiver. In the proposed method, excretion / non-excretion prediction is considered as a two-class classification problem. Here, no excretion data is recorded in the log data, and excretion time is not a true excretion time. Therefore, several temporal parameters are defined for feature extraction. Evaluation experiments of urination / non-urination and defecation / non-defecation were conducted under the conditions of person dependent task and person independent task using four log data recorded in the special nursing home for the elderly. As a result, recognition accuracy of 75% for urination / non-urination and 85% for defecation / non-defecation were obtained and the effectiveness of the proposed method was confirmed.

DOI： 10.23919/SICE.2018.8492590

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85056716876&origin=inward

Image-based Attention Level Estimation of Interaction Scene by Head Pose and Gaze Information 査読有り

Komiya R., Saitoh T., Shimada K.

Proceedings - 17th IEEE/ACIS International Conference on Computer and Information Science, ICIS 2018 497 - 501 2018年09月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. In many situations such as chat and discussion, we interact with other people, and interaction analysis is useful for improving communication skills and evaluating a learning effect of education. Although the audio information is possible to analyze the speaker behavior, it is difficult to analyze the non-speaker. On the other hand, the image information is possible to estimate the inside information, such as the concentration level and understanding level of the interaction of the non-speaker not only the speaker by using nodding, face direction, facial expression, and gaze. This paper proposes an image-based method to automatically estimate the attention level of participants for the interaction scene. We collected simulated interaction scenes by using the omnidirectional camera, and applied the proposed method. As a result, we confirmed the effectiveness of the proposed method.

DOI： 10.1109/ICIS.2018.8466462

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85055721201&origin=inward

Construction of a liar corpus and detection of lying situations 査読有り

Takabatake S., Shimada K., Saitoh T.

Proceedings - 2018 Joint 10th International Conference on Soft Computing and Intelligent Systems and 19th International Symposium on Advanced Intelligent Systems, SCIS-ISIS 2018 971 - 976 2018年07月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. Recognizing lies in human interaction is one of the most important and challenging tasks in artificial intelligence. For example, it leads to the development of a human-computer interaction system that detects a liar in a conversation. In another example, some systems in the future might occasionally need to tell a lie for not hurting user's feelings, namely a white lie for trouble-free conversations. In this paper, we construct a video corpus for developing a system that detects lies in conversations. The corpus consists of 540 question-answer pairs by 9 persons; 270 true answers and 270 lies. We manually analyze the images in the corpus for recognizing lies. In addition, we apply machine learning techniques, support vector machines with local binary pattern features and geometric features, to our corpus. We discuss the results of the analysis of the corpus and the machine learning approaches to our corpus.

DOI： 10.1109/SCIS-ISIS.2018.00161

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85067133291&origin=inward

Analysing a leader’s eye gaze in emergency care simulation 査読有り国際誌

Sakai, T, Nakamura, K, Abe, T, Saitoh, T, Taneichi, A, Tsuchiya, K

International Forum on Quality and Safety in Healthcare 2018 2018年05月

サイレント音声認識

齊藤剛史

知能と情報（日本知能情報ファジィ学会誌） ( 日本知能情報ファジィ学会 ) 30 ( 2 ) 110 - 110 2018年04月

Multimodal analysis of public speaking performance by EFL learners: Applying deep learning to understanding how successful speakers use facial movement 査読有り

Fuyuno M., Komiya R., Saitoh T.

Asian Journal of Applied Linguistics 5 ( 1 ) 117 - 129 2018年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

© 2018, The Asian Journal of Applied Linguistics. Although multimodal corpus analysis has been widely practiced in the field of applied linguistics, few studies have investigated performance of English public speaking by EFL learners. Needs for effective public speaking are fundamental in the globalizing society; however, performing public speaking in English is challenging for EFL learners, and objective analysis on factors of eye contact and speech pauses still remain few though such information is crucial in efficient teaching. This study analyses public speaking performance by EFL learners based on data from a multimodal corpus. Data were collected in an annual speech contest at a Japanese high school. Speakers presented English speeches to an audience and judges. The data consist of video and digital audio recordings of performance, as well as speech scripts and evaluation scores by contest judges. Characteristics of speakers' facial movement patterns in regard to spoken contents and the correlation between facial movements and eye movements were examined. Facial and eye movements were detected with motion tracking and the deep learning method. The results indicated that facial direction changes were not synchronized with speech pauses among highly evaluated speakers. Furthermore, the facial direction changes tended to be synchronized with content words in the spoken utterance rather than function words.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85047367965&origin=inward

サイレント音声認識（Silent Speech Recognition）査読有り

齊藤剛史

知能と情報 ( 日本知能情報ファジィ学会 ) 30 ( 2 ) 110 - 110 2018年01月

Image-based Attention Estimation for Interaction Scene 査読有り

Rinko Komiya, Takeshi Saitoh, and Kazutaka Shimada

International Conference on Information and Communication Technology Robotics (ICT-ROBOT2017) 2017年11月

Kinect sensor based sign language word recognition by mutli-stream HMM 査読有り

Kodama T., Koyama T., Saitoh T.

2017 56th Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2017 2017-November 94 - 99 2017年11月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2017 The Society of Instrument and Control Engineers - SICE. The aim of our research is a development of the sign language recognition (SLR) system for smooth communication between hearing disability people and hearing people. Sign language is consisted of three elements; hand position, hand movement, and hand shape. The proposed method is reflected all elements. Multi-stream HMM is used for the learning and recognition processes. In the experiment, two datasets: 100 Japanese sign language words dataset and 20 Italian gesture dataset, are used for evaluating the proposed method. As a result, all three elements of hand position, hand movement, and LBP-based hand shape are efficient for SLR task.

DOI： 10.23919/SICE.2017.8105627

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044149610&origin=inward

Pupil center detection for infrared irradiation eye image using CNN 査読有り

Kondo N., Chinsatit W., Saitoh T.

2017 56th Annual Conference of the Society of Instrument and Control Engineers of Japan, SICE 2017 2017-November 100 - 105 2017年11月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2017 The Society of Instrument and Control Engineers - SICE. This research proposes a convolutional neural network based pupil center detection method from an eye image which is captured by a wearable eye camera. This paper investigates the detection accuracy by applying several preprocessing, such as edge detection, and binarization. We find the preprocess method in order to speed up the computation time. We collected approximately 8,000 eye images with six subjects, and evaluated the performance, and confirmed that the proposed method obtained higher detection accuracy and have the potential to apply for a real-time application.

DOI： 10.23919/SICE.2017.8105630

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044101536&origin=inward

Intensity Correction Effect for Lip Reading 査読有り

Michiko Kubokawa and Takeshi Saitoh.

28th British Machine Vision Conference (BMVC2017) workshop on Lip-Reading using Deep Learning Methods (LRDLM2017) 2017年09月

Two features combination with gated recurrent unit for visual speech recognition 査読有り

Iwasaki M., Kubokawa M., Saitoh T.

Proceedings of the 15th IAPR International Conference on Machine Vision Applications, MVA 2017 326 - 329 2017年07月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2017 MVA Organization All Rights Reserved. Recently, deep learning has attracted in the visual speech recognition (VSR) field, however, most of approaches use single feature. This paper proposes two features; image-based feature by autoencoder, and motion-based feature based on feature points, and combination approach with gated recurrent unit for VSR task. The proposed method was evaluated on a public dataset Oulu VS, and the speaker independent setting was carried out. Compared with the state-of-the-art, our method obtained the highest accuracy.

DOI： 10.23919/MVA.2017.7986867

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85027835892&origin=inward

Analyzing Public Speaking for EAP Pedagogy: Factors of Better Performance 査読有り

Miharu Fuyuno, Rinko Komiya, Takeshi Saitoh

CAES International Conference Faces of English 2 53 - 53 2017年06月

Fish species recognition based on CNN using annotated image 査読有り

Miyazono T., Saitoh T.

Lecture Notes in Electrical Engineering 449 156 - 163 2017年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© Springer Nature Singapore Pte Ltd. 2018. The objective of our research project is identify the fish species by using image processing technique. This paper proposes a novel feature-points representation method named annotated image. Furthermore, this paper proposes a fish species recognition method based on CNN using proposed annotated image. We collected 50 species of fish images, and applied to the proposed method. As the results, it was confirmed that the annotated image in which all four feature points are plotted into a channel, was obtained the highest recognition accuracy.

DOI： 10.1007/978-981-10-6451-7_19

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85029427359&origin=inward

Head pose estimation using convolutional neural network 査読有り

Lee S., Saitoh T.

Lecture Notes in Electrical Engineering 449 164 - 171 2017年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© Springer Nature Singapore Pte Ltd. 2018. Estimating the head pose is an important capability of a robot when interacting with humans. But there are many difficulties of human head pose estimation, such as extreme pose, lighting, and occlusion, has historically hampered. This paper addresses the problem of head pose estimation with two degrees of freedom (pitch and yaw) using a low-resolution image. We propose a method that uses convolutional neural networks for training and classifying various head poses over a wide range of angles from a single image. Evaluations on public head pose database, Prima database, demonstrate that our method achieves better results than the state-of-the-art.

DOI： 10.1007/978-981-10-6451-7_20

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85029436349&origin=inward

Concatenated frame image based CNN for visual speech recognition 査読有り

Saitoh T., Zhou Z., Zhao G., Pietikäinen M.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 10117 LNCS 277 - 289 2017年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© Springer International Publishing AG 2017.This paper proposed a novel sequence image representation method called concatenated frame image (CFI), two types of data augmentation methods for CFI, and a framework of CFI-based convolutional neural network (CNN) for visual speech recognition (VSR) task. CFI is a simple, however, it contains spatial-temporal information of a whole image sequence. The proposed method was evaluated with a public database OuluVS2. This is a multi-view audio-visual dataset recorded from 52 subjects. The speaker independent recognition tasks were carried out with various experimental conditions. As the result, the proposed method obtained high recognition accuracy.

DOI： 10.1007/978-3-319-54427-4_21

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85016094685&origin=inward

Head pose estimation and motion analysis of public speaking videos 査読有り

Komiya R., Saitoh T., Fuyuno M., Yamashita Y., Nakajima Y.

International Journal of Software Innovation 5 ( 1 ) 57 - 71 2017年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

Copyright © 2017, IGI Global. Public speaking is an essential skill in a large variety of professions and also in everyday life. However, it can be difficult to master. This paper focuses on the automatic assessment of nonverbal facial behavior during public speaking and proposes simple and efficient methods of head pose estimation and motion analysis. The authors collected nine and six speech videos from a recitation and oration contest, respectively, conducted at a Japanese high school and applied the proposed method to evaluate the contestants' performance. For the estimation of head pose from speech videos, their method produced results with an acceptable level of accuracy. The proposed motion analysis method can be used for calculating frequencies and moving ranges of head motion. The authors found that the proposed parameters and the eye-contact score are strongly correlated and that the proposed frequency and moving range parameters are suitable for evaluating public speaking. Thus, on the basis of these features, a teacher can provide accurate feedback to help a speaker improve.

DOI： 10.4018/IJSI.2017010105

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85024120529&origin=inward

CNN-Based Pupil Center Detection for Wearable Gaze Estimation System 査読有り

Chinsatit W., Saitoh T.

Applied Computational Intelligence and Soft Computing 2017 2017年01月

担当区分：最終著者,　責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

© 2017 Warapon Chinsatit and Takeshi Saitoh. This paper presents a convolutional neural network- (CNN-) based pupil center detection method for a wearable gaze estimation system using infrared eye images. Potentially, the pupil center position of a user's eye can be used in various applications, such as human-computer interaction, medical diagnosis, and psychological studies. However, users tend to blink frequently; thus, estimating gaze direction is difficult. The proposed method uses two CNN models. The first CNN model is used to classify the eye state and the second is used to estimate the pupil center position. The classification model filters images with closed eyes and terminates the gaze estimation process when the input image shows a closed eye. In addition, this paper presents a process to create an eye image dataset using a wearable camera. This dataset, which was used to evaluate the proposed method, has approximately 20,000 images and a wide variation of eye states. We evaluated the proposed method from various perspectives. The result shows that the proposed method obtained good accuracy and has the potential for application in wearable device-based gaze estimation.

DOI： 10.1155/2017/8718956

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85042094141&origin=inward

大学講義における小中学校児童生徒向けの理科教室の試み査読有り

齊藤剛史

工学教育 65 ( 5 ) 5_54 - 5_58 2017年01月

担当区分：責任著者記述言語：日本語掲載種別：研究論文（学術雑誌）

This study proposes a new mechanism for training specialists who are conscious of disinterest in science through implementation of science class for elementary and junior high school students in university lecture. This paper summarizes our activities over the five years, and shows the educational significance and results. University students will learn about science education, safety education, and improve their qualities as instructors by not only devising science experiment themes but also implementing science class. Participating students will raise their interest in science, feel university students close to them, promote a wide range of careers, including the go to university.

DOI： 10.4307/jsee.65.5_54

その他リンク： https://ci.nii.ac.jp/naid/130006110349

カーロボ連携大学院での自動車・ロボットの知能化高度化に向けた専門人材育成

森江隆, 石井和男, 我妻広明, 田向権, 榎田修一, 齊藤剛史, 松永良一, 松波勲, 大貝晴俊

工学教育 65 ( 4 ) 4_51 - 4_56 2017年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

The Joint Graduate School Intelligent Car and Robotics Course, called “Car-Robo Joint Graduate School” is an educational program that started in 2012 mainly at The Kitakyushu Science and Research Park (KSRP). Graduate schools of engineering in Kyushu Institute of Technology, the University of Kitakyushu, and Waseda University at KSRP participate in this program. The purpose of this program is to foster highly skilled personnel in technical fields on intelligent car and robot technology. They will gain practical ability as leaders, mastering their special fields, understanding the peripheral technologies, and leading research and development teams. The special-feature program of this course is seven comprehensive practicums as team learning. Internship students of national colleges of technology and other universities as well as master and PhD course students of these graduate schools participate in the practicums.

DOI： 10.4307/jsee.65.4_51

その他リンク： https://ci.nii.ac.jp/naid/130005966874

Kinectを用いた手話認識

齊藤剛史

設計工学 ( 日本設計工学会誌，Vol.51，No.11，pp.760-764，（2016.11）．齊藤剛史．画像処理による割り箸原形の外観検査システムの開発，日 ) 51 ( 11 ) 760 - 764 2016年11月

LBP-TOP based Facial Expression Recognition using Non Rectangular ROI 査読有り

Masaya Iwasaki, Takeshi Saitoh

International Conference on Information and Communication Technology Robotics 2016年09月

CNN for pupil center detection 査読有り

Warapon Chinsatit, Takeshi Saitoh

International Conference on Information and Communication Technology Robotics 2016年09月

Head pose estimation and movement analysis for speech scene 査読有り

Komiya R., Saitoh T., Fuyuno M., Yamashita Y., Nakajima Y.

2016 IEEE/ACIS 15th International Conference on Computer and Information Science, ICIS 2016 - Proceedings 2016年08月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2016 IEEE.Public speaking is an essential skill for a large variety of professions and in everyday life. However, it is difficult to master the skills. This paper focuses on the automatic assessment of nonverbal facial behavior of public speaking and proposes simple and efficient method of head pose estimation and motion analysis. We collected nine speech scenes of the recitation contest in a Japanese high school, and applied the proposed method to evaluate the performance. As for the head pose estimation, our method was obtained acceptable accuracy for the speech scene. Proposed motion analysis method can be calculated frequencies and moving ranges of head motion. As the result, it was found that there is correlation between the moving range and eye contact score.

DOI： 10.1109/ICIS.2016.7550793

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84988019765&origin=inward

Improvement of eye detection performance for inside-out camera 査読有り

Chinsatitf W., Saitoh T.

2016 IEEE/ACIS 15th International Conference on Computer and Information Science, ICIS 2016 - Proceedings 2016年08月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2016 IEEE.This paper presents a fast and precious eye detection technique by using gradient value for improve the performance of inside-out camera. The propose methods using gradient vector to estimate a temporary pupil position before using ellipse fitting to detect actual eye position. We collected 280 eye images that capture from our inside-out camera, and apply the proposed method and previous method to the dataset. The experimental results show that the propose method is improving performance of wearable gaze estimation system.

DOI： 10.1109/ICIS.2016.7550794

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84987958832&origin=inward

Image-based fish recognition 査読有り

Saitoh T., Shibata T., Miyazono T.

Proceedings of the 2015 7th International Conference of Soft Computing and Pattern Recognition, SoCPaR 2015 260 - 263 2016年06月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2015 IEEE. We are studying image-based fish identification. Most of traditional approaches used a fish image which was easy to extract a fish region with a white background or uniform background for automatic processing. This research adapted an approach to give several points by manual operation by the user. The proposed approach is able to accept the fish image in the complicated background taken on the rocky place. Furthermore, to investigate the efficient features for fish recognition, we defined various features, such as, shape features, local features, and six kinds of texture features. We collected 129 species under various photography conditions, and the proposed method was carried out to it. As the results, it was confirmed that a combination features with geometric features and BoVW models obtained the highest recognition accuracy.

DOI： 10.1109/SOCPAR.2015.7492817

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84979258309&origin=inward

Character Input System using Gaze Estimation 査読有り

Warapon Chinsatit, Masataka Shibuya, Kenji Kawada, Takeshi Saitoh

International Conference on Communication Systems and Computing Application Science 2016年03月

Semantic Structure, Spech Units and Facial Movements: Multimodal Corpus Analysis of English Public Speaking 査読有り

Miharu Fuyuno, Yuko Yamashita, Takeshi Saitoh, Yoshitaka Nakajima

International Conference on Corpus Linguistics 2016年03月

D-15-32 スピーチ指導のための顔特徴点を用いた話者頭部の動き解析(D-15.教育工学,一般セッション)

小宮凜子, 齊藤剛史, 冬野美晴, 山下友子, 中島祥好

電子情報通信学会総合大会講演論文集 ( 一般社団法人電子情報通信学会 ) 2016 ( 1 ) 211 2016年03月

Feature points based fish image recognition 査読有り

Saitoh T., Shibata T., Miyazono T.

International Journal of Computer Information Systems and Industrial Management Applications 8 12 - 22 2016年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

© MIR Labs.We are studying image-based fish identification. Most traditional approaches use a fish image wherein extraction is easy given that the fish region contrasts with a white or uniform background. This research introduces an approach to give several feature points by manual operation. In our proposed approach, we are able to accept fish image with complicated background, including rocky area. Further, to investigate efficient features for fish recognition in image, we define various features, including geometric features, bag of visual words (BoVW) models, and texture features. We collected images comprising 129 fish species under various photographic conditions and applied our proposed method to these images. From our results, we confirmed that a combination of geometric features and BoVW models obtained the highest recognition accuracy.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84963527448&origin=inward

Area Type Judgment of Mathematical Document Using Movements of Gaze Point 査読有り

Ryoji Fukuda, Yuki Ogino, Kenji Kawada, Takeshi Saitoh

20th Asian Technology Conference in Mathematics 346 - 353 2015年12月

Optical flow based lip reading using non rectangular ROI and head motion reduction 査読有り

Shiraishi J., Saitoh T.

2015 11th IEEE International Conference and Workshops on Automatic Face and Gesture Recognition, FG 2015 2015年07月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2015 IEEE. The mouth and the lips are deformed by muscles, and such deformations appear on the surface of the skin. Optical flow can be estimated by these deformations. In this study, we investigate optical flow-based lip reading. In natural speech, we slightly move our head, with speech without head movement being restricted. However, traditional optical flow-based approaches have not considered head movement of the speaker. Moreover, movement of the skin varies with facial placement. However, traditional approaches only consider a rectangular region of interest (ROI). This study addresses previous problems and proposes a novel optical flow-based lip reading method that includes head motion and employs facial features based ROIs. We demonstrate that our method exhibits the highest performance among several other methods for lip reading.

DOI： 10.1109/FG.2015.7163160

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84944930950&origin=inward

OKIRAKU search: Leaf images based visual tree search system 査読有り

Saitoh T., Iwata T., Wakisaka K.

Proceedings of the 14th IAPR International Conference on Machine Vision Applications, MVA 2015 242 - 245 2015年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

© 2015 MVA organization. This paper addresses leaf images based visual tree search system called OKIRAKU Search. A user photographs an isolated leaf on a white background, and inputs the photographs to the system. The system automatically extracts the leaf region and computes shape features and color features, and searches it in the known species. The system shows the user the top matched species. The point of our approach is to accept either front leaf image or rear leaf image, or both images. We investigate the effectiveness of the features and images. In our experiment, 297 classes (278 species) are uniquely collected in the fields around our campus. The experimental results reveal that our proposed method obtains high performance. Moreover, we have developed a prototype tree search systems for mobile phone version and web-based version implemented the proposed method.

DOI： 10.1109/MVA.2015.7153176

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84941195066&origin=inward

Applicability of Gaze Points for Analyzing Priorities of Explanatory Elements in Mathematical Documents 査読有り

Ryoji Fukuda, Junki Iwagami, Takeshi Saitoh

Proc. of 19th Asian Technology Conference in Mathematics 2014年11月

チュートリアル講演機械読唇技術 (マルチメディア情報ハイディング・エンリッチメント) 招待有り

齊藤剛史

電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 ( 一般社団法人電子情報通信学会 ) 114 ( 33 ) 29 - 34 2014年05月

担当区分：筆頭著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

機械読唇技術とは,コンピュータを用いて視覚情報より獲得される口唇の動きのみを用いて発話内容を理解する技術である.機械読唇技術は音声認識技術の補助や障害者のコミュニケーション支援など様々な応用が期待されているが,実用レベルにはまだ達していない.本稿では,機械読唇技術に関する基礎技術から公開データベースや発話内容,撮影機器,撮影方向,フレームレートおよび応用などのトピックスを中心に機械読唇技術について解説する.

その他リンク： https://ci.nii.ac.jp/naid/110009902387

画像処理による割り箸原形の外観検査システムの開発

齊藤剛史

画像ラボ ( 日刊工業出版 ) 1 - 7 2014年05月

画像処理技術を用いた樹木同定に関する研究査読有り

齊藤剛史, 岩田聡大

情報処理学会研究報告. CVIM, [コンピュータビジョンとイメージメディア] ( 一般社団法人情報処理学会 ) 2014 ( 14 ) 1 - 8 2014年02月

担当区分：筆頭著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

本論文では画像処理技術を用いた樹木同定に関する研究として葉画像と樹皮画像を用いた二つのアプローチを検討する.葉画像に関しては,先行研究と同様に白背景で撮影した葉表と葉裏の 2 枚の画像を利用する.さらに先行研究を発展させて,基部と葉先の自動検出法および 25 個の特徴量を提案する.278 種 297 クラスの大規模データセットに対して提案手法および従来手法の比較実験を実施した.その結果,第 5 候補までを考慮した場合,95.1 %の高い認識率を得た.一方,樹皮画像に関しては,52 種のデータセットに対して既存のテクスチャ解析手法 6 種を適用した.その結果,LBP が第 5 候補までを考慮した場合,79.3 %の最も高い認識率を得た.二つのアプローチを通じて画像処理技術による樹木同定の可能性を示した.

その他リンク： https://ci.nii.ac.jp/naid/110009673564

画像処理による割り箸原形の外観検査システムの開発査読有り

齊藤剛史

情報処理学会研究報告. CVIM, [コンピュータビジョンとイメージメディア] ( 一般社団法人情報処理学会 ) 2014 ( 15 ) 1 - 5 2014年02月

担当区分：筆頭著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

本研究では割り箸生産における選別工程の機械化を目標とし,画像処理による割り箸原形の外観検査システムの試作機を開発する.本論文で処理対象とする割り箸原形は杉材である.この原形に対して 2 台のカメラで撮影した画像を用いて (A) 目込み・目粗,(B) 皮付き・非皮付き,(C) 半赤・非半赤の 3 項目を検査する手法を提案する.検査項目 (A) に対しては輝度値波形の解析により木目数を計測する手法,検査項目 (B) に対しては HS 色空間において判別分析を適用する手法,検査項目 (C) に対しては輝度値波形の傾きを定量化して判別する手法をそれぞれ提案する.試作機を稼働させて 2000 本の原形画像を撮影して判別実験を行った.その結果,全ての項目において 0.96 以上の高い正解率を確認し提案手法の有効性を確認した.

その他リンク： https://ci.nii.ac.jp/naid/110009673565

指先注視キャリブレーションを利用した注視点推定 (パターン認識・メディア理解) 査読有り

岩上純生, 齊藤剛史

電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 ( 一般社団法人電子情報通信学会 ) 113 ( 431 ) 189 - 194 2014年02月

担当区分：責任著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

本論文では,ユーザが指先を注視することで特別な器具を必要としない指先注視キャリブレーションを利用した注視点推定を提案する.提案手法では独自に開発した2台のカメラ(Eyeカメラ,Sceneカメラ)および赤外線LED群から構成されるInside-Outカメラを用いる.キャリブレーションモードでは,Eyeカメラ画像に対しては瞳孔中心を検出し,Sceneカメラ画像に対しては指先を検出する.注視モードでは,Eyeカメラ画像から検出された瞳孔中心を用いてSceneカメラの注視点を推定する.様々な指先文字の形状や複数被験者に対して注視点推定実験を実施した.その結果,約2.2°の平均視野角度誤差で注視点を推定可能であることを確認した.

その他リンク： https://ci.nii.ac.jp/naid/110009837992

距離画像を用いた単一指文字認識 (パターン認識・メディア理解) 査読有り

井上快, 小山智己, 齊藤剛史

電子情報通信学会技術研究報告 = IEICE technical report : 信学技報 ( 一般社団法人電子情報通信学会 ) 113 ( 431 ) 195 - 199 2014年02月

担当区分：責任著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

指文字は手の形を文字言語に対応させた視覚言語であり,動きを伴う指文字と動きを伴わない指文字がある.従来研究の多くは,動きを伴わない清音の指文字が対象とされていた.本論文では清音だけでなく濁音や半濁音など動きを伴う日本の全指文字を対象として,距離画像を用いた単一指文字認識手法を提案する.提案手法の概要は以下のとおりである.RGB-Dセンサより取得した距離画像より右手領域を抽出する.手領域より特徴量を算出し,HMMを用いて認識する.日本の81指文字を認識対象とし,被験者5名に対して評価実験を実施した.その結果,96%の平均認識率を得て提案手法の有効性を示した.

その他リンク： https://ci.nii.ac.jp/naid/110009837993

Eigenlips using RGB-D Cameras for Lip Reading 査読有り

Yuki Takasaki, Takeshi Saitoh, Tomoki Koyama

20th Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV2014) 180 - 183 2014年02月

Easy Calibration for Gaze Estimation using Inside-Out Camera 査読有り

Junki Iwagami, Takeshi Saitoh

20th Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV2014) 292 - 297 2014年02月

Vanishing point-based road detection for general road images 招待有り査読有り

Bui T., Saitoh T., Nobuyama E.

IEICE Transactions on Information and Systems E97-D ( 3 ) 618 - 621 2014年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

This paper proposes a vanishing point-based road detection method. Firstly, a vanishing point is detected using a texture-based method proposed in a recent study. After that, a histogram is generated for detecting two road borders. The road area is defined as the region between the two road borders and below the vanishing point. The experimental results demonstrate that our method performs well in general road images. Copyright © 2014 The Institute of Electronics, Information and Communication Engineers.

DOI： 10.1587/transinf.E97.D.618

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84897680454&origin=inward

Road area detection based on texture orientations estimation and vanishing point detection 査読有り

Bui T., Saitoh T., Nobuyama E.

Proceedings of the SICE Annual Conference 1138 - 1143 2013年12月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Estimating a proper location of the road area from a single road image plays an essential role in autonomous driving systems and driver assistant systems. In this paper, we propose a new road area detection method based on texture orientations estimation and vanishing point detection. This method first estimates a vanishing point using a texture-based soft voting algorithm proposed in our recent study. After that, texture orientations and color information are combined in order to generate a histogram for estimating two most dominant road borders. The road area is defined as a region between the two detected road borders and below the estimated vanishing point. The proposed method has been imple- mented and tested in 1000 road images which contain large variations in color, texture, lighting condition and surrounding environment. The experimental results demonstrate that our method performs well in challenging conditions.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84888592693&origin=inward

Tree recognition based on leaf images 査読有り

Iwata T., Saitoh T.

Proceedings of the SICE Annual Conference 2489 - 2494 2013年12月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper addresses an image-based tree recognition method using shape features and color features of leaf images. Our method requires two leaf images including a front leaf image and a rear leaf image which are placed on a white background. Firstly, a leaf region is automatically extracted using a graph cuts-based method. Next, sixteen shape features, six color features, and one size feature are calculated from the extracted leaf region. These features are fed to random forests. In our experiments, 96 classes (92 species) with 20 samples are collected in fields around our campus. The experimental results reveal that the recognition accuracy of the rear leaf is higher than that of the front leaf.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84888628912&origin=inward

画像処理による割り箸原形の外観検査システムの開発査読有り

齊藤剛史，今村将訓，福井幸美

電子情報通信学会論文誌 J96-D ( 10 ) 2570 - 2579 2013年10月

画像処理による割り箸原形の外観検査システムの開発 (システム開発論文特集) 査読有り

齊藤剛史, 今村将訓, 福井幸美

電子情報通信学会論文誌. D, 情報・システム ( 一般社団法人電子情報通信学会 ) 96 ( 10 ) 2570 - 2579 2013年10月

担当区分：筆頭著者記述言語：日本語掲載種別：研究論文（学術雑誌）

本研究では割り箸生産における選別工程の機械化を目標とし,画像処理による割り箸原形の外観検査システムの開発を目的とする.本論文で処理対象とする割り箸原形は杉材である.この原形に対して2台のカメラで撮影した画像を用いて(A)目込み・目粗,(B)皮付き・非皮付き,(C)半赤・非半赤の3項目を検査する手法を提案する.検査項目(A)に対しては,輝度値波形の解析により木目数を計測する手法,検査項目(B)に対しては,HS色空間において判別分析を適用する手法,検査項目(C)に対しては,輝度値波形の傾きを定量化して判別する手法をそれぞれ提案する.試作機を稼働させて2000本の原形画像を撮影して判別実験を行った.その結果,全ての項目において0.96以上の高い正解率を確認し,提案手法の有効性を確認した.

その他リンク： https://ci.nii.ac.jp/naid/110009674129

Road Area Detection Based on Texture Orientations Estimation and Vanishing Point Detection 査読有り

Trung Hieu Bui, Takeshi Saitoh, Eitaku Nobuyama

Proc. of SICE Annual Conference 1138 - 1143 2013年09月

Tree Recognition based on Leaf Images 査読有り

Toshiriro Iwata, Takeshi Saitoh

Proc. of SICE Annual Conference 2489 - 2494 2013年09月

Efficient Face Model for Lip Reading 査読有り

Takeshi Saitoh

12th International Conference on Auditory-Visual Speech Processing (AVSP2013) 227 - 232 2013年08月

H-031 葉の表裏を利用した樹木の画像認識(H分野:画像認識・メディア理解,一般論文)

岩田聡大, 齊藤剛史

情報科学技術フォーラム講演論文集 ( FIT(電子情報通信学会・情報処理学会)運営委員会 ) 12 ( 3 ) 161 - 162 2013年08月

H-052 講義映像における移動物体動線の自動検出(H分野:画像認識・メディア理解,一般論文)

西野博貴, 齊藤剛史

情報科学技術フォーラム講演論文集 ( FIT(電子情報通信学会・情報処理学会)運営委員会 ) 12 ( 3 ) 207 - 208 2013年08月

「回路シミュレータQucsによる電子回路入門」（導入編）

齊藤剛史

月刊I/O ( 工学社 ) 38 ( 6 ) 60 - 61 2013年06月

Development of Initial Inspection System for Wooden Chopsticks by Image Processing 査読有り

Takeshi Saitoh, Masanori Imamura, Yukimi Fukui

Proc. of 11th International Conference on Quality Control by Artificial Vision (QCAV2013) 75 - 80 2013年05月

読唇技術の現状と展望

齊藤剛史

電波技術協会報FORN ( 電波技術協会 ) ( 292 ) 32 - 35 2013年05月

Kinect を利用した指文字認識に関する検討

井上快, 齊藤剛史

電気学会研究会資料. MBE, 医用・生体工学研究会 2013 ( 1 ) 45 - 50 2013年01月

Kinectを利用した指文字認識に関する検討 (MEとバイオサイバネティックス)

井上快, 齊藤剛史

電子情報通信学会技術研究報告 : 信学技報 ( 一般社団法人電子情報通信学会 ) 112 ( 417 ) 45 - 50 2013年01月

担当区分：責任著者記述言語：日本語掲載種別：研究論文（研究会，シンポジウム資料等）

本論文ではMicrosoft社製Kinectを用いた指文字認識手法を提案する.Kinectより得られた距離画像から右手領域を自動的に抽出する.次に手領域より指の本数,アスペクト比,外接矩形と手領域の面積比,円形度,距離分布比およびフーリエ記述子の6種類の特徴量を求め,最後にSVMを用いて認識する.リアルタイム処理では,誤認識を防ぐために静止判定および投票処理を導入する.動きを伴わない41種類の日本指文字を認識対象とし,被験者5名に対して評価実験を実施した.その結果,特定人物認識実験では平均認識率95%,不特定人物認識では平均認識率53%の認識率を得た.

その他リンク： https://ci.nii.ac.jp/naid/110009727765

Development of communication support system using lip reading 査読有り

Saitoh T.

IEEJ Transactions on Electrical and Electronic Engineering 8 ( 6 ) 574 - 579 2013年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

This paper develops a complete communication support system for speech and hearing disorders using lip reading. The proposed system is a high-level system and consists of a face detector using the Viola-Jones method, lip extraction based on the active appearance model, dynamic programming matching, automatic utterance section extraction, camera control, and message output. We implemented the system in a commercial laptop, and carried out the recognition experiments on 50 Japanese phrases with four subjects for more than 1 month and obtained high recognition accuracy. Furthermore, we demonstrated it in various situations such as the sitting position, supine position, and holding position. We confirmed that our system can be used in various situations. © 2013 Institute of Electrical Engineers of Japan.

DOI： 10.1002/tee.21898

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84885833913&origin=inward

A texture-based local soft voting method for vanishing point detection from a single road image 査読有り

Bui T., Nobuyama E., Saitoh T.

IEICE Transactions on Information and Systems E96-D ( 3 ) 690 - 698 2013年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

Estimating a proper location of vanishing point from a single road image without any prior known camera parameters is a challenging problem due to limited information from the input image. Most edge-based methods for vanishing point detection only work well for structured roads with clear painted lines or distinct boundaries, while they usually fail in unstructured roads lacking sharply defined, smoothly curving edges. In order to overcome this limitation, texture-based methods for vanishing point detection have been widely published. Authors of these methods often calculate the texture orientation at every pixel of the road image by using directional filter banks such as Gabor wavelet filter, and seek the vanishing point by a voting scheme. A local adaptive soft voting method for obtaining the vanishing point was proposed in a previous study. Although this method is more effective and faster than prior texture-based methods, the associated computational cost is still high due to a large number of scanning pixels. On the other hand, this method leads to an estimation error in some images, in which the radius of the proposed half-disk voting region is not large enough. The goal of this paper is to reduce the computational cost and improve the performance of the algorithm. Therefore, we propose a novel local soft voting method, in which the number of scanning pixels is much reduced, and a new vanishing point candidate region is introduced to improve the estimation accuracy. The proposed method has been implemented and tested on 1000 road images which contain large variations in color, texture, lighting condition and surrounding environment. The experimental results demonstrate that this new voting method is both efficient and effective in detecting the vanishing point from a single road image and requires much less computational cost when compared to the previous voting method. Copyright © 2013 The Institute of Electronics, Information and Communication Engineers.

DOI： 10.1587/transinf.E96.D.690

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84878262881&origin=inward

「はじめてのパターン認識」(平井有三　著) 招待有り

齊藤剛史

計測と制御 ( 公益社団法人計測自動制御学会 ) 52 ( 7 ) 658 - 658 2013年01月

Road Detection Based on an Estimated Vanishing Point and Color Information 査読有り

Trung Hieu Bui, Eitaku Nobuyama, Takeshi Saitoh

World Academy of Science, Engineering and Technology 71 1748 - 1753 2012年11月

Leaf Recognition using Shape and Color Features 査読有り

Takeshi Saitoh, Toshihiro Iwata

World Academy of Science, Engineering and Technology 71 1734 - 1739 2012年11月

Development of inspection system for wooden chopsticks 査読有り

Saitoh T.

Proceedings of the 12th IAPR Conference on Machine Vision Applications, MVA 2011 247 - 250 2011年12月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Our aim is to development of the inspection system of wooden chopsticks, and in this paper, we develop the prototype for high-class disposable wooden chopsticks by computer vision technology. We propose a method of inspecting the four items of fine grain sample, coarse grain sample, unpeeled sample, and defective sample. Moreover, we addressed not only the proposal of the algorithm but development of a prototype system. We carried out the inspection experiment to 1050 samples, and evaluated quantitatively. As the results, for four items of fine grain sample, coarse grain sample, unpeeled sample, and defective sample, 98.5%, 82.2%, 98.9%, and 95.8% of F-measures were obtained.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84872514883&origin=inward

Real-time lip reading system for fixed phrase and its combination 査読有り

Saitoh T.

1st Asian Conference on Pattern Recognition, ACPR 2011 461 - 464 2011年12月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper presents a high-level real-time lip reading system that can recognize both fixed phrase and its combination. Lip reading provides an important means for realizing a communication support interface for speech handicaps. The approach is based on the Viola-Jones face detector, lip extraction based on active appearance model, automatic utterance section detector, and phrase classifier using DP matching. Our system has two unique ideas; combined phrase recognition, and remove false recognition phrase from a target list. These ideas are simple but efficient in practical use. We demonstrated our system with five subjects and confirmed the usefulness. © 2011 IEEE.

DOI： 10.1109/ACPR.2011.6166595

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84862873152&origin=inward

Development of Communication Support System using Lip Reading 査読有り

Takeshi Saitoh

10th International Conference on Auditory-Visual Speech Processing (AVSP2011) 117 - 122 2011年08月

Development of Inspection System for Wooden Chopsticks 招待有り査読有り

Takeshi Saitoh

IAPR Conference on Machine Vision Applications (MVA 2011) 247 - 250 2011年06月

発話シーンからのキーフレーム検出とキーフレームに基づく単語読唇査読有り

齊藤剛史,森下和敏,小西亮介

電気学会論文誌 131 ( 2 ) 418 - 424 2011年02月

Keyframe extraction from utterance scene and keyframe-based word lip reading 査読有り

Saitoh T., Morishita K., Konishi R.

IEEJ Transactions on Electronics, Information and Systems 131 ( 2 ) 418 - 424 2011年01月

記述言語：日本語掲載種別：研究論文（学術雑誌）

In the field of speech recognition, the research on the auditory speech recognition which obtains high recognition accuracy tackles actively. However, in the recognition only using audio information, it is easy to be affected by the influence of surrounding noise, and there is a problem to which use environment is restricted. In recent years, the lip reading attracts attention as one of the interfaces in which use in the scene where the public place and noisy environment place is possible. Many of researches on the conventional lip reading are the word-based approach which recognize based on the information which taken the utterance scene of the word used as the candidate for recognition in advance. In this case, when changing a recognition word, it is necessary to newly take an utterance scene and to re-learn. For this reason, there is a problem which cannot recognize the word which is not registered. Then, in this paper, we propose the new keyframe-based lip reading method which does not need the advanced registration of an utterance scene. © 2011 The Institute of Electrical Engineers of Japan.

DOI： 10.1541/ieejeiss.131.418

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=80052411229&origin=inward

Real-time word lip reading system based on trajectory feature 査読有り

Saitoh T., Konishi R.

IEEJ Transactions on Electrical and Electronic Engineering 6 ( 3 ) 289 - 291 2011年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

This paper describes a real-time word lip reading system. Although our system is based on a word lip reading method already proposed, this system adapts to the user facial movements that cannot be avoided in a real-time process. With nine subjects, we obtained an average recognition rate of 89% for 14 Japanese words. Moreover, we carried out additional experiments with the system under various conditions, and verified the robustness and usefulness of our system. © 2011 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

DOI： 10.1002/tee.20658

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=79953867793&origin=inward

スケジューリングによる待機電力削減機能を持つ多機能コンセントの開発招待有り査読有り

明山寛史,川村尚生,笹間俊彦,菅原一孔,齊藤剛史,小西亮介

情報処理学会論文誌 51 ( 12 ) 2287 - 2297 2010年12月

Profile lip reading for vowel and word recognition 査読有り

Saitoh T., Konishi R.

Proceedings - International Conference on Pattern Recognition 1356 - 1359 2010年11月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper focuses on the profile view, which is the second most typical angle after the frontal face, and proposes a profile view lip reading method. We applied the normalized cost method to detect profile contour. Five feature points, the tip of the nose, upper lip, lip corner, lower lip, and chin, were detected from the contour, and eight features obtained from the five feature points were defined. We gathered two types of utterance scenes, five Japanese vowels and 20 Japanese words. We selected 20 combinations based on the eight features and carried out recognition experiments. Recognition rates of 99% for vowel recognition and 86% for word recognition were obtained with five features: two lip heights, two protrusion lengths, and one lip angle. © 2010 IEEE.

DOI： 10.1109/ICPR.2010.335

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=78149476156&origin=inward

A study of influence of word lip reading by change of frame rate 査読有り

Takeshi Saitoh,Ryosuke Konishi

9th International Conference on Auditory-Visual Speech Processing (AVSP2010) 131 - 136 2010年10月

Current Sensor based Home Appliance and State of Appliance Recognition 査読有り

Takeshi Saitoh,Tomoyuki Osaki,Ryosuke Konishi,Kazunori Sugahara

SICE Journal of Control, Measurement, and System Integration 3 ( 2 ) 86 - 93 2010年03月

Word Lip Reading in Various Tones 招待有り査読有り

Takeshi Saitoh,Hiroyuki Ishikura,Ryosuke Konishi

16th Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV2010) 304 - 308 2010年02月

Voice activated wheelchair with collision avoidance using sensor information 査読有り

Murai A., Mizuguchi M., Nishimori M., Saitoh T., Osaki T., Konishi R.

ICCAS-SICE 2009 - ICROS-SICE International Joint Conference 2009, Proceedings 4232 - 4237 2009年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper develops a functional voice activated wheelchair. Various interfaces to control powered wheelchair are proposed. Since the voice is the most natural communication ways for person, our study pays attention to speech recognition. The user controls the wheelchair by the interactive operation. The wheelchair does not act based on false speech recognition. However, there is a problem that is the wheelchair collides in the wall and obstacle by delaying the voice command. Then, our system applies the collision avoidance function CAF by which wheelchair avoids the wall or obstacle without voice command by using the information of two kinds of sensor. CAF assists the user to control the wheelchair without colliding in the wall or obstacle. The effectiveness of our system was confirmed through the running experiment. © 2009 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=77951112765&origin=inward

Mobile robot with following function and autonomous return function 査読有り

Tsuda N., Harimoto S., Saitoh T., Konishi R.

ICCAS-SICE 2009 - ICROS-SICE International Joint Conference 2009, Proceedings 635 - 640 2009年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper proposes the mobile robot consisted of following function and returning function with monocular camera. In the following function, the robot follows the target object such as the person who walks in front of robot, and moves until reaching his destination. To follow him, the region extraction method is applied. Furthermore, the robot records the moving route. In the returning function, the robot moves by tracing the recorded route. We developed the mobile robot based on the electronic wheelchair and carried out several experiments. As the result, we confirm that our robot moves both following mode and returning mode adequate accuracy. © 2009 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=77951116858&origin=inward

Mobile robot with following and returning mode 査読有り

Tsuda N., Harimoto S., Saitoh T., Konishi R.

Proceedings - IEEE International Workshop on Robot and Human Interactive Communication 933 - 938 2009年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper proposes the mobile robot embedded two functions, the following function and returning function in the indoor environment with monocular camera. In the following mode, the robot follows the target object such as the person who walks in front of robot, and runs until reaching his destination. To follow him, the region extraction method is applied. Furthermore, the robot records the running route. In the returning mode, the robot runs by tracing the recorded route. We developed the mobile robot based on the electronic wheelchair and carried out some experiments. As the result, we confirm that our robot runs both following and returning mode by satisfactory performance. © 2009 IEEE.

DOI： 10.1109/ROMAN.2009.5326252

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=72849143061&origin=inward

Elevator available voice activated wheelchair 査読有り

Murai A., Mizuguchi M., Saitoh T., Osaki T., Konishi R.

Proceedings - IEEE International Workshop on Robot and Human Interactive Communication 730 - 735 2009年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper develops a new functional voice activated wheelchair. Our system has two major functions. The first function is the collision avoidance function that the wheelchair avoids the wall and obstacle without voice command by using the sensor information. This function has three autonomous movements of the stop movement, avoidance movement and deceleration movement. The second function is the robust elevator entry/exit function which can work smoothly in a variety of situations. Various experiments were carried out and evaluate the proposed system with several conditions. As the result, our system ran indoor environment safely. Furthermore, our system was possible to entry/exit the elevator, and move between floors. © 2009 IEEE.

DOI： 10.1109/ROMAN.2009.5326329

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=72849137680&origin=inward

Bird tracking and flapping motion recognition for monitoring system 査読有り

Maruyama K., Saitoh T., Konishi R.

ICCAS-SICE 2009 - ICROS-SICE International Joint Conference 2009, Proceedings 3613 - 3618 2009年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Recently, the increasing bird injury is one of the social problems. To solve this problem, we set our goal to develop the monitoring system. As the first step, we focus the bird tracking and flapping motion recognition both techniques are core technologies of our goal. It is found that Mean Shift tracking based on both the color model and flow model is the most effective to track the target object compared with other well known tracking methods, such as, optical flow, color model based Mean Shift, and Particle Filter. We defined three features the flow gradient and two flow features based on the frequency analysis, and proposed the method for recognizing the flapping motion. We collected 70 scenes including the bird's flying, bird's walking, and person's walking. As the result, the flow gradient feature is the effective for recognizing the flapping motion. © 2009 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=77951102480&origin=inward

A novel transducer: From lip motion to voice message 査読有り

Saitoh T., Kato T., Konishi R.

Proceedings of the 11th IAPR Conference on Machine Vision Applications, MVA 2009 410 - 413 2009年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes the novel transducer which recognizes lip motion of the word utterance and output the recognized word as the voice message. Though, the base method of our transducer is speaker dependent word lip reading, our system is adopted the facial movement of the user. For the developed prototype system, the robustness was evaluated to the two kinds of facial movement, the distance between the user and camera changed and rotation movement of the face. As the result, we confirmed the robustness above movements. Moreover, the recognition experiment with three subjects was carried out for two word groups of 14 words and 47 words. The recognition rates of 90.0% and 82.6% were obtained with 14 words group and 47 words group, respectively.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84862905261&origin=inward

Mobile Robot with Following and Returning Mode 招待有り査読有り

Naoki Tsuda,Shuji Harimoto,Takeshi Saitoh,Ryosuke Konishi

18th IEEE International Symposium on Robot and Human Interactive Communication (IEEE RO-MAN2009) 933 - 938 2009年09月

Elevator Available Voice Activated Wheelchair 招待有り査読有り

Akira Murai,Masaharu Mizuguchi,Takeshi Saitoh,Tomoyuki Osaki,Ryosuke Konishi

18th IEEE International Symposium on Robot and Human Interactive Communication (IEEE RO-MAN2009) 730 - 735 2009年09月

Voice Activated Wheelchair with Collision Avoidance Using Sensor Information 招待有り査読有り

Akira Murai,Masaharu Mizuguchi,Masato Nishimori,Takeshi Saitoh,Tomoyuki Osaki,Ryosuke Konishi

ICROS-SICE International Joint Conference 2009 (ICCAS-SICE09) 4232 - 4237 2009年08月

Mobile Robot with Following Function and Autonomous Return Function 招待有り査読有り

Naoki Tsuda,Shuji Harimoto,Takeshi Saitoh,Ryosuke Konishi

ICROS-SICE International Joint Conference 2009 (ICCAS-SICE09) 635 - 640 2009年08月

Bird Tracking and Flapping Motion Recognition for Monitoring System 招待有り査読有り

Kohei Maruyama,Takeshi Saitoh,Ryosuke Konishi

ICROS-SICE International Joint Conference 2009 (ICCAS-SICE09) 3613 - 3618 2009年08月

単眼カメラを用いた中央走行型の屋内移動ロボット査読有り

齊藤剛史,多田直也,小西亮介

電気学会論文Ｃ 129 ( 8 ) 1576 - 1583 2009年08月

A Novel Transducer: From Lip Motion To Voice Message 査読有り

Takeshi Saitoh,Tomoya Kato,Ryosuke Konishi

IAPR Conference on Machine Vision Applications (MVA 2009) 410 - 413 2009年05月

Indoor mobile robot navigation by central following based on monocular vision 査読有り

Saitoh T., Tada N., Konishi R.

IEEJ Transactions on Electronics, Information and Systems 129 ( 8 ) 2009年01月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper develops the indoor mobile robot navigation by center following based on monocular vision. In our method, based on the frontal image, two boundary lines between the wall and baseboard are detected. Then, the appearance based obstacle detection is applied. When the obstacle exists, the avoidance or stop movement is worked according to the size and position of the obstacle, and when the obstacle does not exist, the robot moves at the center of the corridor. We developed the wheelchair based mobile robot. We estimated the accuracy of the boundary line detection, and obtained fast processing speed and high detection accuracy. We demonstrate the effectiveness of our mobile robot by the stopping experiments with various obstacles and moving experiments. © 2009 The Institute of Electrical Engineers of Japan.

DOI： 10.1541/ieejeiss.129.1576

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=72549100735&origin=inward

Analysis of Efficient Lip Reading Method for Various Languages 査読有り

Takeshi Saitoh,Kazutoshi Morishita,Ryosuke Konishi

19th International Conference on Pattern Recognition (ICPR2008) 2008年12月

Analysis of efficient lip reading method for various languages 査読有り

Saitoh T., Morishita K., Konishi R.

Proceedings - International Conference on Pattern Recognition 2008年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

The traditional researches targeted at only one language, and there is no research to refer the language and recognition method. Moreover, a lot of model-based methods use only an external lip or intraoral region, and tooth or tongue region is not reflected to the feature. This paper describes analysis of efficient lip reading method for various languages. First, we applies active appearance model, and simultaneously extracts the external and internal lip contour. Then, the tooth and intraoral regions are detected. Various features from five regions are fed to the recognition process. We set four languages to be the recognition target, and recorded twenty words per each language. As the result, proposed trajectory feature based on three shape features, the area and aspect ratio of internal lip region, and area of intraoral region, was obtained the highest recognition rates of 93.6%, compared with the traditional methods and other regions. © 2008 IEEE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=77957930895&origin=inward

Japanese 45 Single Sounds Recognition Using Intraoral Shape 査読有り

Takeshi Saitoh,Ryosuke Konishi

IEICE Information and Systems E91-D ( 11 ) 2735 - 2738 2008年11月

Japanese 45 single sounds recognition using intraoral shape 査読有り

Saitoh T., Konishi R.

IEICE Transactions on Information and Systems E91-D ( 11 ) 2735 - 2738 2008年11月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper describes a recognition method of Japanese single sounds for application to lip reading. Related researches investigated only five or ten sounds. In this paper, experiments were conducted for 45 Japanese single sounds by classifying them into five vowels category, ten consonants category, and 45 sounds category. We obtained recognition rates of 94.7, 30.9 and 30.0% with trajectory feature. Copyright © 2008 The Institute of Electronics, Information and Communication Engineers.

DOI： 10.1093/ietisy/e91-d.11.2735

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=68749103384&origin=inward

Current Sensor based Non-intrusive Appliance Recognition for Intelligent Outlet 査読有り

Takeshi Saitoh,Yuuki Aota,Tomoyuki Osaki,Ryosuke Konishi,Kazunori Sugahara

23rd International Technical Conference on Circuits/Systems, Computers and Communications（ITC-CSCC2008） 349 - 351 2008年07月

Monocular Vision based Indoor Mobile Robot 査読有り

Naoya Tada,Keisuke Murata,Takeshi Saitoh,Tomoyuki Osaki,Ryosuke Konishi

23rd International Technical Conference on Circuits/Systems, Computers and Communications（ITC-CSCC2008） 41 - 44 2008年07月

Voice controlled intelligent wheelchair 査読有り

Nishimori M., Saitoh T., Konishi R.

Proceedings of the SICE Annual Conference 336 - 340 2007年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

In order to assist physically handicapped persons, we developed a voice controlled wheelchair. The user can control the wheelchair by voice commands, such as "susume (run forward)" in Japanese. A grammar-based recognition parser named "Julian" is used in our system. Three type commands, the basic reaction command, the short moving reaction command, and the verification command, are given. We experimented speech recognition by Julian, and obtained a successful recognition rate 98.3%, 97.0% of the movement command and the verification command, respectively. The running experiment with three persons was carried out in the campus room and the utility of our system is shown. © 2007 SICE.

DOI： 10.1109/SICE.2007.4421003

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=50249089982&origin=inward

Oral motion controlled intelligent wheelchair 査読有り

Saitoh T., Takahashi N., Konishi R.

Proceedings of the SICE Annual Conference 341 - 346 2007年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes the novel interface equipped intelligent wheelchair. The proposed interface is using visual oral motion. In our system, the user moves his mouth open and close to run the wheelchair. To achieve detection of mouth region and analysis mouth motion in real-time, we proposed mouth cavity region detection method. We carried out the operational experiment and the running experiment with five persons, and obtained proposed system brings the user toward destination operated with oral motion. © 2007 SICE.

DOI： 10.1109/SICE.2007.4421004

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=50249097141&origin=inward

Mobile robot navigation by center following using monocular vision 査読有り

Tada N., Saitoh T., Konishi R.

Proceedings of the SICE Annual Conference 331 - 335 2007年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

A system for center following in the corridor using monocular vision is presented. The robot detects a vanishing point and obstacle region from frontal view image. The vanishing point is found by two boundary lines of the wall and the corridor. The obstacle region is detected from corridor region using color histogram. The experiments are performed on the mobile robot. The results show that the robot runs the center of the corridor and stops when an obstacle is detected within the target range from 1m to 2m distance. © 2007 SICE.

DOI： 10.1109/SICE.2007.4421002

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=50249164125&origin=inward

Japanese phone recognition using lip image information 査読有り

Saitoh T., Hisagi M., Konishi R.

Proceedings of IAPR Conference on Machine Vision Applications, MVA 2007 134 - 137 2007年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes Japanese phone recognition for lip reading based on a novel feature called trajectory feature to obtain high recognition rate. Trajectory feature is a time change of two mouth shape features expressed as a two-dimensional trajectory of the lip motion. The most similar trajectory in a database which is compared with the target trajectory by DP matching is selected as a result phone. Traditional researches concerned phone recognition experimented with only Japanese five vowels. In this paper, experiments were conducted for Japanese 45 phones with five person, and classified into five vowels and ten consonants, and recognition rate of 94.1% and 28.9%, respectively, were obtained.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84872588189&origin=inward

Development of an intelligent wheelchair with visual oral motion 査読有り

Saitoh T., Takahashi N., Konishi R.

Proceedings - IEEE International Workshop on Robot and Human Interactive Communication 145 - 150 2007年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes an intelligent wheelchair with a novel interface which uses visual oral motion. The wheelchair runs in various lighting environments, traditional mouth region detection method did not consider these conditions. To achieve detection of mouth region and analysis mouth motion in real-time and in various lighting environments, we proposed mouth cavity region detection method. In our system, the user moves his mouth open and close to run the wheelchair. We carried out six operational experiments and running experiments. Experiments with five subjects obtained our system is supposed to use a certain extent in a campus corridor. ©2007 IEEE.

DOI： 10.1109/ROMAN.2007.4415070

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=48749087074&origin=inward

Analysis of Features for Efficient Japanese Vowel Recognition 査読有り

Takeshi Saitoh,Mitsugu Hisagi,and Ryosuke Konishi

IEICE Information and Systems E90-D ( 11 ) 1889 - 1891 2007年11月

Analysis of features for efficient Japanese vowel recognition 査読有り

Saitoh T., Hisagi M., Konishi R.

IEICE Transactions on Information and Systems E90-D ( 11 ) 1889 - 1891 2007年11月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper analyses the features required to efficiently recognize five Japanese vowels for lip-reading. Various features, such as shape and radius, are calculated from the lip region and fed to the k Nearest Neighbor method. We calculated 15 feature sets and found that the feature set including the area and aspect ratio of the mouth cavity is effective for Japanese vowel recognition. Copyright © 2007 The Institute of Electronics, Information and Communication Engineers.

DOI： 10.1093/ietisy/e90-d.11.1889

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=68249144673&origin=inward

Oral Motion Controlled Intelligent Wheelchair 査読有り

Takeshi Saitoh,Noriyuki Takahashi,Ryosuke Konishi

SICE Annual Conference 2007，International Conference on Instrumentation, Control and Information Technology 341 - 346 2007年09月

Mobile Robot Navigation by Center Following using Monocular Vision 査読有り

Naoya Tada,Takeshi Saitoh,Ryosuke Konishi

SICE Annual Conference 2007，International Conference on Instrumentation, Control and Information Technology 331 - 335 2007年09月

Voice Controlled Intelligent Wheelchair 査読有り

Masato Nishimori,Takeshi Saitoh,Ryosuke Konishi

SICE Annual Conference 2007，International Conference on Instrumentation, Control and Information Technology 336 - 340 2007年09月

Development of an Intelligent Wheelchair with Visual Oral Motion 査読有り

Takeshi Saitoh,Noriyuki Takahashi,Ryosuke Konishi

16th IEEE International Symposium on Robot and Human Interactive Communication (IEEE RO-MAN2007) 145 - 150 2007年08月

オプティカルフロー分布を利用したMean Shift追跡　色モデルとフローモデルを統合したMean Shift追跡法の提案

小西亮介,齊藤剛史,大島直也

画像ラボ ( 日刊工業出版 ) 52 - 56 2007年08月

Japanese Phone Recognition using Lip Image Information 査読有り

Takeshi Saitoh,Mitsugu Hisagi,Ryosuke Konishi

IAPR Conference on Machine Vision Applications (MVA 2007) 134 - 137 2007年05月

トラジェクトリ特徴量に基づく単語読唇査読有り

齊藤剛史,小西亮介

電子情報通信学会論文誌 J90-D ( 4 ) 1105 - 1114 2007年04月

オプティカルフロー分布を利用したMean Shift追跡査読有り

大島直也,齊藤剛史,小西亮介

電子情報通信学会論文誌 J90-D ( 4 ) 1096 - 1104 2007年04月

Real time mean shift tracking using optical flow distribution 査読有り

Oshima N., Saitoh T., Konishi R.

2006 SICE-ICASE International Joint Conference 4316 - 4320 2006年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes about real-time object tracking method based on mean shift tracking algorithm. The mean shift tracking algorithm is an efficient technique for tracking object through an image. The original mean shift tracking is proposed to apply the color image based on the color distribution. A Near-Infrared Camera is used with surveillance system to take in the dark. It is difficult to track the target object in the low contrast image such as the infrared image. To overcome this problem, our idea is to consider optical flow distribution. The proposed method is integrated three distributions (color, flow magnitude and flow direction). Experiments were conducted for the color image and the infrared image compared with the original method. It will be shown that our method will be able to track a target object in the low contrast image. © 2006 ICASE.

DOI： 10.1109/SICE.2006.314969

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=34250723920&origin=inward

Monocular autonomy following vehicle 査読有り

Saitoh T., Osaki T., Konishi R.

2006 SICE-ICASE International Joint Conference 5963 - 5966 2006年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes the development of the monocular autonomy following vehicle which for reduce a helper burden. The system is composed by a CCD camera, a FPGA board and some control circuits. Our idea is that the distance between the system and the target human is related to the size of the human region in the camera image. Based on this idea, we proposed the automatic human region detection method with S-ACM which is a kind of Active Contour Model. We implemented our method and developed the system. The experiment which followed the target human was done. As the result, we confirmed that the system followed with keeping a fixed distance with the target human. © 2006 ICASE.

DOI： 10.1109/SICE.2006.315838

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=34250777309&origin=inward

Lip reading using video and thermal images 査読有り

Saitoh T., Konishi R.

2006 SICE-ICASE International Joint Conference 5011 - 5015 2006年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes the lip reading method using video and thermal images. We pay attention to the lip movement. The density of the thermal image expresses the surface temperature, and it is difficult to detect the lip position automatically. Though the video and the thermal images take in the same time, these don't synchronized in time and spatially. Then the time matching and the spatial matching process are applied. Next, a lip region is detected from the video image, and a lip region of the thermal image is detected corresponding to the video region. Finally, the eigenimage waveform is computed, and DP matching is applied. Experiments were conducted for 75 scenes (five words of three subjects). A successful recognition rate of 76.0% with the video image, 44.0% with the thermal image, and 80.0% with both images were obtained. Experimental results show the recognition rate was improves by combining with the video and the thermal images. © 2006 ICASE.

DOI： 10.1109/SICE.2006.315095

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=34250757596&origin=inward

A real-time footstep tracking for monitoring system 査読有り

Saitoh T., Kodani T., Konishi R.

Proceedings of the 8th IASTED International Conference on Signal and Image Processing, SIP 2006 403 - 408 2006年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

When a person is walking the outside of the room, the location of the person is unknown with the camera image because the wall is occluded the person and the camera. This paper proposed a real-time system for tracking a footstep based on direction of arrival estimation method. The microphone array technique is applied to estimate the direction of the target person and the camera with a motor is turned to the person in real-time. We applied simple estimation method with 1125 footstep sounds and obtained 80.3% of the success rate with cross-correlation method of input signals. The tracking demonstration was carried out with five examinees. Though the recording time was limited only a half second because of the hardware problem, we confirmed that the direction of the examinee who walks a corridor could be detected.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=56549097136&origin=inward

回転可動カメラによるリアルタイム人物位置推定システム査読有り

大島直也,齊藤剛史,小西亮介

電気学会論文誌C 126 ( 8 ) 957 - 962 2006年08月

Word recognition based on two dimensional lip motion trajectory 査読有り

Saitoh T., Konishi R.

2006 International Symposium on Intelligent Signal Processing and Communications, ISPACS'06 287 - 290 2006年01月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

The visual information doesn't have an influence with the noise in comparison with the auditory information. However, lip-reading is earlier technology than speech recognition, there is a problem that recognition rate is low. This paper proposes a novel feature to obtain high recognition rate for lip-reading. The proposed lip-reading method is as follows. First, Sampled Active Contour Model is applied to detect lip contour during the input image sequences. The lip size is then normalized. Next, a time change of two lip shape features (area and aspect ratio) is expressed as a two-dimensional trajectory of the lip motion. The most similar trajectory in a database which is compared with the target trajectory by DP matching is selected as a result word. We recorded ten Japanese words that uses for the wheelchair control. As a result of comparative experiment with other features in which has already been proposed, trajectory feature obtained the highest recognition rate of 97.2%. © 2006 IEEE.

DOI： 10.1109/ISPACS.2006.364888

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=45249101841&origin=inward

Real-time person localization system with an active camera 査読有り

Ooshima N., Saitoh T., Konishi R.

IEEJ Transactions on Electronics, Information and Systems 126 ( 8 ) 957 - 962 2006年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper proposed a method in which location of a person is estimation with one active camera in real-time. The approach taken in this paper is as follows. First, we detect the person region with proposed method which is the region detection method based on short time distribution of the density to avoid the influence of the camera movement. Then we obtain his height and the gravity point. The rotate angle and the distance between the person and the camera are calculated by the information of the height and the gravity point. Experiments were applied with several examples. We obtained that the average error distance is less than 12[cm]. The average error distance is small enough in comparison with shoulder width of the persons. Our method can estimate the location in real-time.

DOI： 10.1541/ieejeiss.126.957

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=33746904743&origin=inward

Automatic moving object detection and tracking with mean shift for surveillance system 査読有り

Oshima N., Saitoh T., Konishi R.

2006 International Symposium on Intelligent Signal Processing and Communications, ISPACS'06 578 - 581 2006年01月

記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes an automatic detection and tracking method for unknown object from unclear background. Our method consists of two processes, the detection process and the tracking process. The former process is the background subtraction method and the block-based clustering. The latter process is with mean shift method. To detect an accurate size of the target region, we set a correction period. Demonstration was carried out with a scene which three men are appeared and walked. It is found that two regions could be tracked even when two men was walked toward different direction which they walking together. We confirmed that the proposed method could process on real time. © 2006 IEEE.

DOI： 10.1109/ISPACS.2006.364724

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=45249105798&origin=inward

Analysis of efficient feature for Japanese vowel recognition 査読有り

Hisagi M., Saitoh T., Konishi R.

2006 International Symposium on Intelligent Signal Processing and Communications, ISPACS'06 33 - 36 2006年01月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

Recently, speech, especially word recognition using visual information, has attracted significantly interest. In word recognition, the target is not just a word, but also a vowel. However, since vowel frame do not contain many phonemes, vowel recognition rate is less than that of the word. Some research has been done on vowel recognition. This paper analyses features that can effectively recognize five Japanese vowels. The process of our method is as follows. First, Sampled Active Contour Model is applied to detect lip contour during the input image sequences. The lip size is then normalized. Next, various features, such as the shape, diameter, and the approximate function features, are calculated and compared with the recognition rates using k Nearest Neighbor method. Two experiments, speaker dependent recognition and speaker independent recognition are carried out with five subjects. We calculated eleven feature sets and found that the feature set including the area and aspect ratio is the most effective. In the speaker dependent recognition, we obtained the recognition rate of 96.6%, and in the speaker independent recognition, we obtained 82.4%. © 2006 IEEE.

DOI： 10.1109/ISPACS.2006.364829

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=45249108717&origin=inward

自然画像に基づく花画像の自動認識査読有り

齊藤剛史,金子豊久

電子情報通信学会論文誌 J89-D-II ( 12 ) 2341 - 2349 2005年12月

Automatic speaker detection based on voice and image processing 査読有り

Saitoh T., Ooshima N., Konishi R.

Proceedings of the SICE Annual Conference 3368 - 3371 2005年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

In this paper, we propose a small-scale remote video conference system, and this paper describes a method for detecting a speaker among the participants based on the voice and the image processing of lip. At first, its method detects the direction of sound source and turns to camera to a speaker's direction. Then, the lip shape which has the most active, is detected as a speaker. We experimented to detect a speaker from some scenes. As a result, we obtained the successful rate of 98% at permissible rate angle of ±15°. And we obtained the successful rate of 100% by the image processing. © 2005 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=33645304320&origin=inward

Route search method by normalized cost 査読有り

Saitoh T., Kaneko T.

Systems and Computers in Japan 36 ( 14 ) 11 - 20 2005年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

In this paper, we propose a method for extracting an object region from an image. First, we investigate the properties of the Intelligent Scissors (IS) method, which has been proposed as a method for extracting contours. As a result, we verify that the IS method's route search according to total cost requires multiple seed points to extract a complex contour. Next, we show that this problem can be resolved by using the route length to normalize the total cost and then minimizing the normalized cost. Based on this principle, we propose a contour extraction method based on the normalized cost, which we term the Normalized Cost (NC) method. In addition, we construct a user interface that includes the NC method and perform quantitative comparison experiments. From the experiments comparing the NC method with the IS method, we verify that the NC method can extract a region faster and with fewer seed points than the IS method. We use these results to show the effectiveness of the proposed technique. © 2005 Wiley Periodicals, Inc.

DOI： 10.1002/scj.20420

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=27844555330&origin=inward

Real-time invader tracking system with surveillance camera 査読有り

Ooshima N., Saitoh T., Konishi R.

Proceedings of the SICE Annual Conference 3364 - 3367 2005年12月

担当区分：責任著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

In this paper, we proposed an automatic extraction method of the contour image of an invader and the automatic calculation method of tracks with one surveillance camera image. For extraction method of region, we measured the density value distribution using past several flames every each pixel and proposed the method which judges whether moving object exists or not. Also, for the track of the invader, we obtain the relation between the height of moving object and distance in advance, and measured the position from the height of image. Moreover, we calculated the angle from the position in the image of invader and invader's actual position is calculated. Simulation experiment was done by using the track of one person who walked in the one room. As a result, the effectiveness of the proposal technique was shown. © 2005 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=33645308535&origin=inward

Lip reading based on sampled active contour model 査読有り

Saitoh T., Konishi R.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 3656 LNCS 507 - 515 2005年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes a model-based method for detecting lip region from image sequences. Our approach is by Sampled Active Contour Model (S-ACM). The original S-ACM has the problem which can't expand. To overcome this problem, we propose the elastic S-ACM. Moreover, based on the extracted lip contour, the effective delta radius features are led to the word HMM. We recorded ten words that uses for the wheelchair control, and obtained a recognition rate of 89% with twelve features. © Springer-Verlag Berlin Heidelberg 2005.

DOI： 10.1007/11559573_63

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=33646001053&origin=inward

Lip detection and visual speech recognition 査読有り

Saitoh T., Fukui Y., Konishi R.

Proceedings of the SICE Annual Conference 2951 - 2954 2005年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes an automatic method for detecting lip region by Sampled Active Contour Model (S-ACM). The original S-ACM has the problem which can't expand. In this paper, to overcome this problem, we proposed the elastic S-ACM. Moreover, by using this S-ACM, we detected lip contour, and the bimodal recognition system based on visual obtained from lip contour and audio information is proposed. Twelve delta radius features from image sequence and sixteen delta cepstrum features from audio data are fed to a word HMM. We recorded ten words that uses for wheelchair control. We obtained bimodal recognition rates of approximately 95%, 93%, and 89% if SNR=30dB, 15dB and 0dB, respectively. © 2005 SICE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=33645287546&origin=inward

Optimal threshold for hepatic blood vessels based on structural analysis and cancer detection 査読有り

Saitoh T., Ikeda M., Aoki K., Kaneko T., Sekiguchi R.

Systems and Computers in Japan 36 ( 7 ) 1 - 12 2005年06月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

Recently, high-resolution multislice CT images have been used in the diagnosis of hepatic cancer. In the conventional method of detecting hepatic cancer on the basis of three-dimensional images, it is difficult to detect cancers with a three-dimensional structure. This paper proposes a method of detection of liver cancer on the basis of three-dimensional hepatic blood vessel regions which are extracted by threshold processing. First, the liver entrance is located by tracing the blood vessels from the abdominal aorta. In order to detect the cancer, the hepatic vessel region must be extracted accurately. For this purpose, a method is proposed in which a temporary threshold is determined near the liver entrance, and the structure of the blood vessel is analyzed by adjusting the threshold from the temporary value in order to determine the optimal threshold. In the structural analysis, thinning is applied to the blood vessel and a directed graph is constructed. When the threshold is too low, a loop is produced due to overextraction of the blood vessel region. The optimal threshold is determined on the basis of this property. The cancer detection procedure is composed of two stages. Primary detection finds cancers contained in the extracted blood vessel region, and secondary detection finds cancers from the region that is not extracted. In experiments on seven cases, ten cancers were detected with three false positives. © 2005 Wiley Periodicals, Inc.

DOI： 10.1002/scj.20329

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=18744369937&origin=inward

自己組織化マップを用いたハーブ種の判別査読有り

尾崎知幸,大島直也,齊藤剛史,小西亮介

計測自動制御学会論文集 41 ( 5 ) 383 - 387 2005年05月

Automatic recognition of blooming flowers 査読有り

Saitoh T., Aoki K., Kaneko T.

Proceedings - International Conference on Pattern Recognition 1 27 - 30 2004年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（国際会議プロシーディングス）

This paper describes an automatic method for recognizing a blooming flower based on a photograph taken with a digital camera in natural scene. The problem of identifying an object against the background is known to be difficult. In this paper, we employ a photograph where the object (a blooming flower) is focused but the background is defocused. For extracting a flower region, we propose a new method that extracts a boundary by selecting a route with minimizing a sum of the local cost divided by the route length. Experiments were conducted for 600 pictures (20 pictures each for 30 species). A successful boundary extraction rate of 97% and a flower recognition rate of 90% were obtained.

DOI： 10.1109/ICPR.2004.1333997

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=10044299384&origin=inward

正規化コストに基づく輪郭抽出法査読有り

齊藤剛史,金子豊久

電子情報通信学会論文誌 J87-D-II ( 12 ) 2136 - 2144 2004年12月

CT画像における構造解析に基づく肝臓内血管の抽出と肝臓癌の検出査読有り

齊藤剛史,池田雅敏,青木公也,金子豊久,関口隆三

電子情報通信学会論文誌 J87-D-II ( 6 ) 1359 - 1368 2004年06月

Automatic Segmentation of Liver Region Based on Extracted Blood Vessels 査読有り

Saitoh T., Tamura Y., Kaneko T.

Systems and Computers in Japan 35 ( 5 ) 1 - 10 2004年05月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

With the rapid progress of CT (Computed Tomography) scanners, four-phase CT images with resolutions as high as 1 mm have started to be used for diagnosing liver diseases. The first-, second-, third-, and fourth-phase CT images correspond to before dye injection, the early stage, the full stage, and the wash-out stage of the injected dye. Such CT data offer useful information for diagnosing hepatic cancer. This paper describes an automatic method for segmenting the liver region using the blood vessel stream in the first- and third-phase CTs by tracing the portal vein and then the hepatic vein. First, the tracing of the veins is carried out by performing a 3D labeling operation in the region extracted with a threshold that separates blood vessels from liver soft tissue. The adjoining stomach and spleen regions are separated by applying erosion and dilation, which are 3D morphological operations. Then, an operation for determining the accurate liver region is carried out to derive an approximate liver region, by enlarging the liver blood vessel region by morphological dilation and then extracting the liver region with an optimal threshold. This method was applied to eight CT data. The resulting regions agree well with the manually detected regions. © 2004 Wiley Periodicals, Inc.

DOI： 10.1002/scj.10592

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=2342433763&origin=inward

Transient Texture Synthesis Based on Multiple Templates 査読有り

Takeshi Saitoh,Manabu Sato,Toyohisa Kaneko,Shigeru Kiriyama

MGV: Machine GRAPHICS and VISION journal 12 ( 4 ) 525 - 537 2004年04月

Automatic extraction of object region from photographs 査読有り

Saitoh T., Aoki K., Kaneko T.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2749 1130 - 1137 2003年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper presents a new method for automatically extracting an object region from a photograph based upon a well-known method "Intelligent Scissors" (IS). For our application, it will be shown that (1) the cost should not be based on accumulated cost adopted by IS but rather on average cost and (2) only a few past pixels are needed for deciding the future route. It will be shown that our new method will be able to extract object region with a correct rate of approximately 93% if object images are well-focused. © Springer-Verlag 2003.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=35248814434&origin=inward

Automatic recognition of wild flowers 査読有り

Saitoh T., Kaneko T.

Systems and Computers in Japan 34 ( 10 ) 90 - 101 2003年09月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

In this paper we propose an automatic recognition system for wild flowers. Two photos, one of the flower and one of the leaves taken from directly above or at a close angle of a single wild flower, were used as a single set. The objective (flower, leaf) is extracted from each image using a clustering method, then recognition is performed using a piecewise linear discriminant function after finding 10 features in the picture of a flower and 11 features in the picture of a leaf. We performed experiments on 20 sets of 34 species of wild flowers that grow around our university campus in the spring and early summer. The results of the experiment showed a recognition rate of 96.0% when all of the features (21) were used. Next we performed an experiment to select the features particularly useful for recognition. These results showed that six features for the pictures of flowers and two features for the pictures of leaves were the most effective, and with them a recognition rate of 96.8% could be reached.

DOI： 10.1002/scj.10099

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=0042420467&origin=inward

血管形状に基づく肝臓領域の自動抽出査読有り

齊藤剛史,田村雄太,金子豊久

電子情報通信学会論文誌 J86-D-II ( 5 ) 633 - 641 2003年05月

Automatic segmentation of liver region through blood vessels on multi-phase CT 査読有り

Saitoh T., Tamura Y., Kaneko T.

Proceedings - International Conference on Pattern Recognition 16 ( 1 ) 735 - 738 2002年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

Recently four phases CT has been widely used for diagnosing liver diseases, especially cancer. This paper describes an automatic method for segmenting the liver region from third phase abdominal CT. First, blood vessels in the liver are extracted with a threshold. Then morphological dilation enables us to define an approximate liver region, which is found to be very useful for removing abutting organs. The final liver region is extracted with a threshold. Experiments were conducted for eight CT data. The extracted regions agree well with those detected manually. © 2002 IEEE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=18744416865&origin=inward

花と葉による野草の自動認識査読有り

齊藤剛史,金子豊久

電子情報通信学会論文誌 J84-D-II ( 7 ) 1419 - 1429 2001年07月

Automatic recognition of wild flowers 査読有り

Saitoh T., Kaneko T.

Proceedings - International Conference on Pattern Recognition 15 ( 2 ) 507 - 510 2000年12月

担当区分：筆頭著者記述言語：英語掲載種別：研究論文（学術雑誌）

This paper describes an automatic method for recognizing wild flowers. Recognition requires two pictures, a frontal flower image and a leaf image taken by a digital camera. Seventeen(17) features, eight from the flower and also nine from the leaf, are fed to a neural network. We collected 20 pairs of pictures from 16 wild flowers in the fields around our campus. We obtained a recognition rate of 95% with all the 17 features. Then we investigated which features are more effective for recognition and found that four features of flowers and two features of leaves can yield the best accuracy of 96%. © 2000 IEEE.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=0043098025&origin=inward

　前のページ-次のページ　

1