研究者詳細 - 中藤　良久

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85089724027&origin=inward

Speaker verification method using htm for security system 査読有り

Sakaguchi Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Advances in Intelligent Systems and Computing 1253 AISC 160 - 165 2021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）

In recent years, we have been using biometric verification systems in various places such as daily life and businesses. However, it’s insufficient in hospital and food factory to introduction of the security system of the room access control. This is because they wear gloves, masks and hats in hospitals and factories, so they cannot verify faces or fingerprints. In this study, we propose a speaker verification system using HTM (Hierarchical Temporal Memory). We verify speaker and detection of unlearned speakers by experiments. In addition, we consider the use in hospitals and factories, and also evaluate it in a noisy environment. The evaluation values are Precision, Recall, and F-measure. The experimental results show that in the case of 10 learners, F-measure exceeded 0.6 in the classification of 2 classes of learners and non-learners.

DOI： 10.1007/978-3-030-55307-4_25

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85089626810&origin=inward

Effectiveness of dynamic image features for estimating daily conversation atmosphere 査読有り

Natsumi K., Wakita Y., Nakatoh Y.

IEEE International Conference on Artificial Intelligence in Engineering and Technology, IICAIET 2020 2020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）

To achieve a conversation atmosphere estimation method, we confirm the usefulness of optical flow values for conversation atmosphere estimation. We extracted only the response scenes from a database of free conversation between elderly individuals, analyzed two-dimensional optical flow parameters, and calculated the standard deviation values (FX-SD, FY-SD) of all frames for each dimension. These values express the dynamic characteristics of conversation gestures. As a result of comparing the FX-SD values and the FY-SD values of 'smooth conversation scene (not-need help (NNH))' with that of 'not-smooth conversation scene (need help (NH))', the 'NNH scenes' values are smaller than those of the 'NH scenes'. We also confirmed the identification rate using support vector machine (SVM). The average identification rate is 72.3%. These results suggest that optical flow parameters are useful for identification, but the performance is not sufficient. In the future, we should evaluate more conversation databases and provide more in-depth discussion of the effectiveness.

DOI： 10.1109/IICAIET49801.2020.9257834

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098051857&origin=inward

Speech Enhancement System Using Lip-reading 査読有り

Matsui K., Fukuyama K., Nakatoh Y., Kato Y.O.

IEEE International Conference on Artificial Intelligence in Engineering and Technology, IICAIET 2020 2020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）

We have been developing a practical speech enhancement system that supports for laryngectomee. By interviewing users we captured essential issues, such as 'utilization of existing device', 'the appearance needs to be inconspicuous', and 'the device should be easy to use'. Considering those user's needs, we plan to use smart phone platform and develop speech enhancement application so that the users are just ordinary looking, and there is no need to buy any additional device. In order to realize such system, the key concept of our proposed system performs lip-reading and speech synthesis. In this study, we examined a lip-reading method that can recognize by registering the words that you want to speak and that is optimized for the user using a small amount of data. 36 viseme images were converted into very small data using VAE(Variational Auto Encoder), then the training data for word recognition model was generated. Viseme is a group of phonemes with identical appearance on the lips. Our viseme sequence representation with VAE was used to be able to adapt users with very small amount of training data set. Word recognition experiment using VAE encoder and CNN was performed with 20 Japanese words. The experimental result showed 65% recognition accuracy, and 100% including 1st and 2nd candidates. The lip-reading type speech enhancement seems appropriate for embedding mobile devices in consideration of both usability and small vocabulary recognition accuracy.

DOI： 10.1109/IICAIET49801.2020.9257839

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098077233&origin=inward

Programming assistant system using improved flowchart for the visually impaired 査読有り

Hirakawa R., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Advances in Intelligent Systems and Computing 1018 496 - 501 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（学術雑誌）

© 2020, Springer Nature Switzerland AG. The purpose of this study is to assist the visually impaired to develop applications. There is a method for the visually impaired to develop applications by reading program codes with screen reader. In this method, the visually impaired can read code on line by line and select a line by knocking the up and down arrow key on the keyboard. After that, they can read the processing and describe an instruction before and after the processing. However, such an operation takes them a long time. Therefore, we suggested the system which let develop applications by editing flowcharts. This system converts program code into flowchart and read blocks in the flowchart. In this system, they can select blocks by knocking the left, right, up and down arrow key. We compared the time for rewriting programs between approach and the proposed system. As a result, the time reduced in all programs.

DOI： 10.1007/978-3-030-25629-6_77

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85070016135&origin=inward

Software Log Anomaly Detection Through One Class Clustering of Transformer Encoder Representation 査読有り

Hirakawa R., Tominaga K., Nakatoh Y.

Communications in Computer and Information Science 1224 CCIS 655 - 661 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（学術雑誌）

© 2020, Springer Nature Switzerland AG. For smart devices such as smartphones and tablets, developing new software using open source software (OSS) is becoming mainstream. While OSS-based development can greatly increase project productivity, it is more difficult to identify the cause of software defects. In this paper, we propose a deep learning model that performs unsupervised learning based on the log data accumulated in the project and calculates the degree of abnormality per line for newly given logs. The proposed method is evaluated using open supercomputer system log data, Blue Gene/L, and the accuracy of the proposed method is compared with the conventional log anomaly detection method using LSTM AutoEncoder. As a result of the comparative experiment, it was found that the proposed method performed better than the conventional method in the two scores of AUROC and F1 Score at the cutoff point.

DOI： 10.1007/978-3-030-50726-8_85

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85088749659&origin=inward

Study on Software Log Anomaly Detection System with Unsupervised Learning Algorithm 査読有り

Hirakawa R., Tominaga K., Nakatoh Y.

Advances in Intelligent Systems and Computing 1152 AISC 122 - 128 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（学術雑誌）

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020. In recent years, the existence of open source software (OSS) is indispensable for software development. While developer can benefit from functions of OSS, there is a problem that it is very difficult to locate the cause when problems occur. In this study, we propose a method to calculate anomaly score for each line of log data. In our method, the temporal pattern is learned using Hierarchical Temporal Memory, which is an unsupervised real-time learning algorithm, and the anomaly score is obtained based on the internal state of the model. In the experiment, we compare the learning situation in the following three input formats, word ID, word embedding, and sentence embedding. In the experiments using actual log data, it was found that the method with word ID has the highest f1 score and runtime performance, but the precision needs to be improved in order to suppress useless information.

DOI： 10.1007/978-3-030-44267-5_18

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85085196359&origin=inward

Discussion on differences in frequency band required for horizontal sound localization by age 査読有り国際誌

Okuno T., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. In this paper, the horizontal sound image experiment of the real sound source was performed under the head resting condition using the band - limited noise using the low pass filter and the band pass filter. Then, the effect of each frequency band on direction perception of the hearing person and the elderly person was clarified. Five people with healthy men in their twenties and five elderly people in their 70s male hearing white noise pass through low-pass filters (2, 3, 4, 5, 6 kHz) and band-pass filters (1-4, 2-4, 2-5 [kHz]) passed through the directional perception experiment. As a result, similarly to the previous study [1], it is difficult to perceive directions sufficiently if two or more frequency bands are included, and it is difficult to perceive the direction of band-pass noise through a band-pass filter of a narrow frequency band. As a result of this research, Direction perception was found that the frequency band around 4 [kHz] and 6 [kHz] is small as an element of direction perception and the frequency band around 5 [kHz] is an important clue for direction perception.

DOI： 10.1145/3325291.3325395

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075439420&origin=inward

Study on watching system for door-to-door sales detection and risk determination 査読有り国際誌

Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. In recent years, malicious visit sales are increasing in Japan, targeting elderly people living alone and dementia patients. Since it is difficult for them to deal with problems with their own power, it is necessary for surrounding people to support damage deterrence. So, we aim to realize a watching system that anyone can use if there is only a smart speaker. In this study, we propose a method to detect door-to-door sales from conversation information of elderly people and visitors with the aim of realizing a system to prevent elderly consumers' troubles. In the evaluation experiment, we evaluate how accurately door-to-door sales can be detected out of the conversation dataset. In addition, we try to determine the risk degree by learning the door-to-door sales sample data given the risk label.

DOI： 10.1145/3325291.3325391

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075470875&origin=inward

Study on method of veering correction by vibration transmission in walking support for visually impaired people 査読有り国際誌

Hirayama T., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. In Japan, accidents are occurring when visually impaired people go out. For example, a collision with an obstacle such as a person or a wall, or a fall from the station platform. A current measure for this problem is a voice guidance. However, visually impaired people place importance on auditory information. Therefore, it is difficult for them to hear surrounding sounds during listening the voice guidance. There is a previous research that trying to guide the visually impaired people to their destination by vibration transmission. However, although this study have realized the guidance of the visually impaired people to their destination without interrupting environmental sounds, this method cannot correct the veering in real time. For above reason, we propose a veering correction system by vibration transmission, which does not disturb acquisition of auditory information. This is a system that gives them vibration by wearing a vibration motor on their shoulders and corrects the veering. In this article, we conducted verification experiment of the effectiveness of the veering correction by vibration transmission as a preliminary study in the development of the veering correction system. As a result of the research, we were able to keep the deviations from over 100 cm to less than 100 cm. As a conclusion, vibration transmission is effective as a method of correcting the veering of visually impaired people.

DOI： 10.1145/3325291.3325392

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075481449&origin=inward

Study on evaluating risks of routes damages at Earthquakes for evacuation guidance system 査読有り国際誌

Mihara Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. In recent years, the damage caused by the earthquake has become very large in Japan. Secondary disasters such as tsunami and fire are the main causes of death due to the earthquake, both of which are causes of evacuation delay. Rapid evacuation is important in the event of a disaster. However, due to the earthquake, various obstacles will occur in the aisle. Obstacles may occur due to the collapse of houses due to the earthquake and it may take time to pass. If you can obtain information about the aisle in advance and avoid evacuation routes in advance, you can complete evacuation safer and more quickly. As a method of acquiring traffic information, We thought that the use of SNS is useful from the viewpoint of the real-time nature of information and the penetration rate of young people aggressively actively communicating information. In this research, we proposed an evacuation guidance system to evacuate users with a safe shortest route. Then we examined the danger of passing through obstacles. As a result, "road shredder" is evaluated as the most dangerous, and "road crack" is evaluated as the most secure.

DOI： 10.1145/3325291.3325390

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075445253&origin=inward

In load collapse prevention system Examination of the relationship between truck vibration and load inclination 査読有り国際誌

Inoshita C., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. ACM In recent years, the demand for the logistics industry has increased enormously due to the spread of Internet mail order. On the other hand, the number of fatal accidents caused by trucks is more than 10,000 a year in Japan. Above all, there is a problem that there is an annual fatal and fatal accident due to truck load collapse. In this study, using the drive recorder, we calculated the inclination of the package photographed during driving by using KCF of object tracking method of image processing. Comparing the change of the Luggage tilt and the amount of change in 3-axis acceleration attached to the vehicle/ time of each frame in all the data, frames at each maximum value point almost agreed at the time of” When riding over convex part of inclined road”. On the other hand, “When a sudden steering wheel during driving”, a point where the absolute value of the change in package inclination during one frame reaches 0.005 ° and the point where the maximum change occurs after that. It almost coincided with the point where the curve was reached in the video. As for the 3-axis acceleration sensor, there is no difference in tendency between the curve and the straight running, and it is considered as the main cause that the influence of the roughness of the road is greatly influenced rather than the influence at the time of the curve.

DOI： 10.1145/3325291.3325393

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075447390&origin=inward

Examination of acoustic features for discriminating between real and loudspeaker speeches 査読有り国際誌

Kinoshita Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery. In the train while driving, there are many noises such as noises from the vibration of the vehicle and the shock noise generated at the vehicle connection, and the roar of human beings, etc. In addition to stressing the passengers, these noises it prevents you from listening to the announcements in the train. In this paper, a new noise canceling earphone system aimed at improving in-car announcement speech listening and reducing stress that the noise gives to passengers by separating surrounding speech and in-train announcement speech regarded as noise I devised. In this study, as a preliminary step to complete the system, in order to find the difference between the ambient speech and the announcement speech such as the roar of the passenger, both speeches were compared using acoustic features. As a result, among the acoustic features used in the comparison, differences were found in the spectral centroid, the spectral roll-off, and the intensity and spectral contrast when divided into sub-bands.

DOI： 10.1145/3325291.3325394

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075462372&origin=inward

Enhancement of electrolarynx speech based on WaveRNN 査読有り国際誌

Urabe E., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

ACM International Conference Proceeding Series 2019年05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）

© 2019 Association for Computing Machinery.ACM In Japan, there are about 30,000 people whose larynxes have been removed. These people are not able to make voice, so they have problems in communication. For example, they cannot communicate their will enough with communication by writing or gestures alone, thereby avoiding communication. However, they can move the mouth and tongue normally. Therefore, they speak using alternative vocalizations. Electrical artificial larynx (EL) is one of the alternative vocalizations. When the vibrator is pressed against the lower part of the jaw, the vibration sound is transmitted in the sound path. This speech method is easy to learn and is cleaner than other speech methods, but this voice sound is bad quality because it has no intonation and it is noisy. In this study, we tried to improve the EL sound quality by using WaveRNN that is one of Deep learning. After making WaveRNN learn human voice, we converted the EL voice. We evaluated the converted voice in a subjective evaluation experiment in 10 words that selected each consonant and vowels in a balanced manner. The evaluation items are: 1. Clarity of consonants, 2. Clarity of vowels, 3. Clarity of whole voice, 4. Natural, 5. Easy to hear. As a result, Clarity and naturalness of consonants improved in words containing two or more unvoiced consonants. On the other hand, in the voiced consonants and vowels, no improvement was seen by the proposed method. Therefore, we think that voice quality was improved in unvoiced consonants. On the other hand, there was no tendency for improvement in voiced consonants and vowels.

DOI： 10.1145/3325291.3325396

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85075454294&origin=inward

Preliminary study of mobile device-based speech enhancement system using lip-reading 査読有り国際誌

Matsunaga Y., Matsui K., Nakatoh Y., Kato Y., Lopez-Sanchez D., Rodriguez S., Corchado J.

Advances in Intelligent Systems and Computing 800 308 - 315 2019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）

© Springer International Publishing AG, part of Springer Nature 2019. Inconspicuous speech enhancement system for laryngectomies using lip-reading is proposed to improve the usability and the speech quality. The proposed system uses a tiny camera on mobile phone and recognize the vowel sequences using lip-reading function. Three types of Japanese vowel recognition algorithms using MLP, CNN, and MobileNets, were investigated. 3,000 image datasets for training and testing were prepared from five persons while uttering discrete vowels. Our preliminary experimental result shows that the MobileNets is appropriate for embedding mobile devices in consideration of a performance both recognition accuracy and calculation cost.

DOI： 10.1007/978-3-319-94649-8_37

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85049977480&origin=inward

Estimating Age-Dependent Degradation Using Nonverbal Feature Analysis of Daily Conversation 査読有り国際誌

Kana N., Wakita Y., Nakatoh Y.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 11582 LNCS 222 - 231 2019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）

© 2019, Springer Nature Switzerland AG. In this paper, we study a system that estimates the degree of decline in the driving ability of elderly people using non-verbal information from daily conversations. It is necessary for us to find the factors that would affect the calculation of the degree of decline that has reached a problematic level for functioning daily life. We focus on the cases where elderly people cannot understand their partner’s speech as their hearing and concentration abilities decrease with age. We analyze the relationship between the degree of understanding of the partner’s speech and the non-verbal characteristic of the response scene. Based on the results of the acoustic analysis of each utterance, the fundamental frequency (F0) and acoustic power levels of when a person can understand their partner’s speech tend to be higher than those when they cannot. The analysis of the synchronization of the head motions shows that brightness value of difference image when a person can understand their partner’s speech is also higher than when they cannot. These results indicate that these non-verbal factors are effective in estimating the decline in the hearing and concentration abilities of the elderly.

DOI： 10.1007/978-3-030-22219-2_17

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85069857489&origin=inward

Potential for Estimating Age-Dependent Degradation using the Acoustic Characteristics of Daily Conversation 査読有り国際誌

Wakita Y., Kondo S., Nakatoh Y.

2018 IEEE 7th Global Conference on Consumer Electronics, GCCE 2018 760 - 763 2018年12月

記述言語：英語掲載種別：研究論文（学術雑誌）

© 2018 IEEE. Our aim is to create a system to estimate the degree of age-dependent degradation in elderly persons based on their daily conversations. We have already confirmed that the time-dependent features of the fundamental frequency (F0) are effective for estimating conversation smoothness. Here we investigated whether these features are also effective for estimating age-dependent degradation. By analyzing free conversational speech data, we confirmed that the standard deviation value of F0 for each utterance is correlated to the average value of F0 when the impression of the speech is regarded as under 75. This correlation was weaker in older individuals. Our results indicate that the above features of F0 would be effective for estimating age-dependent degradation.

DOI： 10.1109/GCCE.2018.8574840

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060313749&origin=inward

音声スペクトルの時間変化強調による明瞭性改善手法の提案査読有り

川原　竣介、平川　凜、中藤　良久

産業応用工学会論文誌 6 ( 1 ) 51 - 54 2018年03月

Development of hands-free speech enhancement system for both EL-users and esophageal speech users 査読有り国際誌

Matsunaga Y., Matsui K., Nakatoh Y., Kato Y.

Advances in Intelligent Systems and Computing 620 334 - 341 2018年01月

記述言語：英語掲載種別：研究論文（学術雑誌）

© Springer International Publishing AG 2018. A hands-free speech enhancement system for laryngectomies is proposed to improve the usability and the speech quality. The system has a small and right weight transducer, photo-reflector based lip movement level sensors, and a microphone. The sensor outputs are used for generating the transducer turn on/off signal. A simple performance test was conducted if the system can determine the utterance segments accurately, and the result showed good utterance detection performance. Also, we tested with five more subjects and confirmed that the proposed method was well performed even under the speaker-independent condition.

DOI： 10.1007/978-3-319-62410-5_41

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85022188286&origin=inward

Method for Improving Intelligibility of Elderly Speech Based on Formant Frequency Shift 査読有り

TANAKA YUTO, MIZUMACHI MITSUNORI, NAKATOH YOSHIHISA

Journal of the Institute of Industrial Applications Engineers ( The Institute of Industrial Applications Engineers ) 5 ( 1 ) 32 - 36 2017年01月

Study on the improvement of intelligibility for elderly speech using formant frequency shift method 査読有り国際誌

Tanaka Y., Mizumachi M., Nakatoh Y.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9811 683 - 690 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）

© Springer International Publishing Switzerland 2016. In general, aging is progressing in developed countries. Elderly people have difficulty controlling their articulation accurately due to aging. We need to improve the quality of elderly speech for smooth communication. In this paper, we analyzed the 1st formant frequency (F1) and 2nd formant frequency (F2) between the more intelligible speech and less intelligible speech of Japanese elderly people. In addition, we improved the intelligibility of less intelligible elderly speech by using the formant frequency shift method. This method is the correcting by shift value of formant frequency based on LPC. The shift value is the magnification such as expanding the F1-F2 size of less intelligible speech.

DOI： 10.1007/978-3-319-43958-7_83

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84984829622&origin=inward

LPC残差駆動歌唱システムにおける音韻性除去による音質改善の検討査読有り

姫野剛至, 田中裕人, 水町光徳, 松井謙二, 中藤良久

産業応用工学会論文誌 ( 一般社団法人産業応用工学会 ) 4 ( 1 ) 19 - 25 2016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）

We propose a singing system to transmit the pseudo-vocal sound source into the vocal tract. In this paper, we examine the pseudo-vocal sound source that is suitable for singing system. So we carried out a comparison of LPC residual wave and Bark-LPC residual wave that the phonological component included in pseudo-vocal sound source was removed by Bark spectrum. As a result, we found that although the singing voice using Bark-LPC residual is in the low natural tendencies, the Bark-LPC residual wave can reproduce a higher speaker information contributed by the speaker characteristics.

DOI： 10.12792/jjiiae.004.01.04

その他リンク： http://ci.nii.ac.jp/naid/130005263509

Singing System by Electric-Larynx Using LPC Residual Wave 査読有り国際誌

Takashi Himeno, Yuto Tanaka, Mitsunori Mizumachi, Yoshihisa Nakatoh, Kenji Matsui

Journal of the Institute of Industrial Applications Engineers 3 ( 4 ) 183 - 186 2015年12月

Study of Hearing Aid System Adapted to Direction Inputted by User 査読有り

NOGUCHI TSUBASA, MIZUMACHI MITSUNORI, NAKATOH YOSHIHISA

Journal of the Institute of Industrial Applications Engineers ( The Institute of Industrial Applications Engineers ) 3 ( 2 ) 72 - 79 2015年04月

文字入力を用いた発話支援システムの開発査読有り

濱井裕也, 水町光徳, 中藤良久, 松井謙二

産業応用工学会論文誌 3 ( 1 ) 10 - 17 2015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）

Utterance is very important for us to communicate with others. However, many people have speech handicaps and can't produce utterance smoothly. They mainly rely on writing and sign language for communication. But they are not easy skills for them to master. In order to help them, we developed a new utterance support system which can function through an information terminal. We made predictions which includes two additional functions of fixed phrase dictionary and the class 2-gram – one of the statistical approaches for processing natural language. Fixed phrase dictionary includes phrases often used in our daily life conversations and class 2-gram uses co-occurrence frequencies of parts of speech in Japanese. There are two purposes for us to use class 2-gram, one is to generate next candidates and the other is to change frequency. We concluded two evaluation experiments. The results show that our system needs less numbers of inputs in helping with written utterance production and using our prototype permits us to input more characters in one minute than using iPhone IME. Hereafter, we intend to create more effective dictionaries and develop new GUI to make our system more effective and faster.

DOI： 10.12792/jjiiae.003.01.02

その他リンク： http://ci.nii.ac.jp/naid/130005263536

Study of Improvement of Intelligibility for Elderly Speech Based on Formant Frequency Shift 査読有り国際誌

Yuto Tanaka, Hiroaki Igaue, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Journal of Computer, Consumer and Control 3 ( 3 ) 57 - 65 2014年08月

音声認識応用システムのための音声コマンド設計ガイドライン査読有り

花沢健, 中藤良久, 中嶌信弥

電子情報通信学会誌 = The journal of the Institute of Electronics, Information and Communication Engineers ( 一般社団法人電子情報通信学会 ) 97 ( 7 ) 582 - 587 2014年07月

記述言語：日本語掲載種別：研究論文（学術雑誌）

近年,音声認識技術とそれを応用したシステムは急速に身近になってきている.音声認識応用システムでは,設計段階でアプリケーションの機能とそれに対応する音声コマンドを定義する必要がある.音声コマンドの設計は,開発者にとってはシステムの性能を引き出す重要事項であり,利用するユーザにとっては使い勝手を決定する主要因となる.JEITA音声入出力方式標準化専門委員会では,音声コマンドの親密度や正確性,発声や音声認識のしやすさなど,音声コマンドの設計における留意点や指針を整理し,ガイドラインとしてまとめている.本稿では,そのエッセンスについて解説する.

その他リンク： https://ci.nii.ac.jp/naid/110009830844

The guidelines for design of voice commands in application systems with speech recognition 査読有り国際誌

Hanazawa K., Nakatoh Y., Nakajima S.

Journal of the Institute of Electronics, Information and Communication Engineers 97 ( 7 ) 582 - 587 2014年01月

Study of Utterance Support System using Word Input based on Word Prediction 査読有り

HAMAI YUYA, MIZUMACHI MITSUNORI, NAKATOH YOSHIHISA

Journal of the Institute of Industrial Applications Engineers ( 産業応用工学会 ) 2 ( 1 ) 13 - 18 2014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）

Speech is very important for us to communicate with others. However, because of speech handicaps, there are many people who feel it difficult. Thus, we set our main purpose to study new utterance support system working on an information terminal. To make this system more effective, we made word prediction and added fixed phrase dictionary and class 2-gram as additional functions. Fixed phrase dictionary includes phrases often used in daily life conversations and class 2-gram using co-occurrence frequencies of 9 parts of speech in Japanese. And by using class 2-gram, we made two additional functions. One is generating next candidates and another is changing frequency. We did three evaluation experiments – necessary numbers of input, input speed and measurement experiments of voice synthesis. And results shows that our system needs less numbers of input but slower than other methods and it is faster to do voice synthesis when user input a sentence than other timing. Hereafter, we intend to create more effective dictionaries, improve class 2-gram and develop new GUI to make our system more effective and faster.

その他リンク： https://ci.nii.ac.jp/naid/120006870840

Speech enhancement by non-linear beamforming tolerant to misalignment of target source direction 査読有り

HAYASHI HIROAKI, MIZUMACHI MITSUNORI, NAKATOH YOSHIHISA

Journal of the Institute of Industrial Applications Engineers ( The Institute of Industrial Applications Engineers ) 1 ( 2 ) 97 - 104 2013年10月

Development of Hand Gesture Controlled Electrolarynx 査読有り

MATSUI KENJI, NAKATOH YOSHIHISA, KATO YUMIKO O.

Journal of the Institute of Industrial Applications Engineers ( The Institute of Industrial Applications Engineers ) 1 ( 2 ) 48 - 51 2013年10月

A low-bit-rate audio codec using mel-scaled linear predictive analysis 査読有り

Nakatoh Y., Matsumoto H.

Acoustical Science and Technology 28 ( 3 ) 147 - 152 2007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）

In this paper, we propose a low-bit-rate audio codec using a new analysis method named mel-scaled linear predictive analysis (mel-LP analysis). In mel-LP analysis, a spectral envelope is estimated on a mel- or bark-frequency scale, so as to improve the spectral resolution in the low-frequency band. This analysis is accomplished with about a twofold increase in computation over standard LPC analysis. Our codec using mel-LP analysis consists of five key parts: time frequency transformation, flattening of MDCT coefficients using the mel-LP spectral envelope, power normalization, perceptual weighting estimation, and multistage VQ. In subjective experiments, we investigated the performance of our codec using the mel-LP analysis method, through 7-level paired comparison tests. The result shows that the codec using the mel-LP analysis method results in a good performance at a low bit rate, particularly at 16 kbps. In the cases of pop songs, piano music and male speech, sound quality was improved. © 2007 The Acoustical Society of Japan.

DOI： 10.1250/ast.28.147

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=34247642253&origin=inward

Speech recognition interface system for digital TV control 査読有り

Nakatoh Y., Kuwano H., Kanamori T., Hoshimi M.

Acoustical Science and Technology 28 ( 3 ) 165 - 171 2007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）

In this paper, we describe a speech recognition interface system for digital TV (DTV) control. TV systems are currently undergoing digitalization and will become more multifunctional, leading to more complex TV operations. Thus, it is necessary for everyone to be able to use TVs easily, and a speech recognition interface is an important key technology. A speech recognition system, which is designed for home use, particularly for digital TV, must be simple and robust to environmental noises and speaker variations. To provide robustness to noise, we developed a noise reduction technique for house noise and an echo-canceling technique for TV sound. To achieve robustness to speaker variations, we developed new speaker adaptation techniques which are incorporated in the system. These of technologies results in a significant improvement in the recognition performance of the DTV. © 2007 The Acoustical Society of Japan.

DOI： 10.1250/ast.28.165

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=34247639555&origin=inward

Generation of broadband speech from narrowband speech based on linear mapping 査読有り

Nakatoh Y., Tsushima M., Norimatsu T.

Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi) 85 ( 8 ) 44 - 53 2002年08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, a method for generating broadband speech from band-limited speech using spectral linear mapping is proposed. This method is based on LPC analysis and synthesis. It first extracts sound source information (residual waveforms) and vocal tract information (spectral envelopes) from input speech, performs linear mapping of the vocal tract information and nonlinear processing of the sound source information for broadband conversion, and finally generates broadband speech from both by LPC synthesis. The spectral envelope is made broadband by linear mapping, the spectral space is divided into a number of subspaces, and a narrowband spectrum is converted into a broadband spectrum by a conversion (convert) matrix of each subspace. A conversion (convert) matrix is estimated using training speech such that the mean square error between the postconversion spectrum and the target broadband spectrum is minimized. Spectral distortions of the proposed method were compared experimentally with codebook mapping and neural network methods and it has been verified that the proposed method with linear mapping gives performance not inferior to that of the other two methods. In addition, the proposed method has been verified to have the effect of giving a band feeling in subject evaluation tests. © 2002 Wiley Periodicals, Inc. Electron. Comm. Jpn. Pt.

DOI： 10.1002/ecjb.10065

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=0036687129&origin=inward

スペストル線形写像による帯域制限音声の広帯域化査読有り

中藤良久, 津島峰生, 則松武志

電子情報通信学会論文誌. D-2, 情報・システム 2-パターン処理 ( 一般社団法人電子情報通信学会 ) 83 ( 11 ) 2246 - 2254 2000年11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）

本論文では, スペクトル線形写像を用いて帯域制限された音声から広帯域音声を生成する方法について提案している.本方法はLPC分析合成を基本とし, まず入力音声から声道情報(スペクトル包絡)と音源情報(残差波形)を抽出し, 声道情報は線形写像により, また音源情報は非線形処理を施すことで広帯域化する.最後に, この両者からLPC合成により広帯域化音声を作成している.線形写像によるスペクトル包絡の広帯域化は, スペクトル空間をいくつかの部分空間に分割し, 各部分空間ごとの変換行列により狭帯域スペクトルから広帯域スペクトルへと変換する.変換行列は, 変換後のスペクトルと目標となる広帯域スペクトルとの間の平均2乗誤差を最小化するように学習用音声を用いて推定される.実験では, 本手法のほかにコードブック写像及びニューラルネットを用いる方法についてスペクトルひずみに関して比較した結果, 線形写像による方法は他の2手法に劣らない性能が得られることが確認できた.また主観評価実験によっても, 帯域感を付与する効果が確認できた.

その他リンク： https://ci.nii.ac.jp/naid/110003183835

Estimation of Road Damage Contents for Earthquake Evacuation Guidance System 査読有り国際誌

Yujiro Mihara, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of 3rd International Conference on Communication and Computational Technologies (ICCCT 2021) 2021年02月

Speech Enhancement System using SVM for Train Announcement 査読有り

Kinoshita Y., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2021-January 2021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）

On the train, announcement speech is one of the important information to know the operation status. However, clarity of announcement speech is not enough because announcement speech is masked by noise. Therefore, new system is proposed to improve clarity of announcement speech based on sound classification. By classifying several sounds, users can select the announcement speech only they want to hear. In this paper, classifier is developed that distinguishes announcement speech using SVM. In the experiments, this classifier obtain that the accuracy of developed classifier is 96%.

DOI： 10.1109/ICCE50685.2021.9427600

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85106052569&origin=inward

Eye Fatigue Prediction System Using Blink Detection Based on Eye Image 査読有り

Kuwahara A., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2021-January 2021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）

Today, information devices such as smartphones have become indispensable in our daily life. As a result, the increased use of information devices has led to the accumulation of eye strain and the development of various eye diseases. As the accumulation of eye-fatigue is manifested in the change in the number of blinks, blink detection technology is necessary. In the previous paper, we proposed a highly accurate blink detection method called EARM. However, there is a problem with this technique, which is that the noise generated by the movement of the face can reduce the accuracy of blink detection. In this study, we used face image normalization to improve the blink detection system by reducing the noise associated with face movement. We found that the proposed system outperformed the results of blink detection accuracy using EARM in all subjects.

DOI： 10.1109/ICCE50685.2021.9427681

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85106066793&origin=inward

Detecting System Alzheimer's Dementia by 1d CNN-LSTM in Japanese Speech 査読有り

Nishikawa K., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2021-January 2021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）

Dementia patients have increased year by year in many developed countries. We need to solve this problem. The questionnaire method is generally used for present dementia screening. However, this method requires time for 10 to 15 minutes with a doctor and a clinical psychologist, which puts a burden on hospitals and test subjects. For this reason, we propose a detecting method using deep learning by means of 1d CNN-LSTM to construct a dementia discriminant system using speech voice. Our method obtained the Accuracy of 0.908 and the F-measure of 0.897.

DOI： 10.1109/ICCE50685.2021.9427692

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85106001096&origin=inward

Blink detection using image processing to predict eye fatigue 査読有り

Kuwahara A., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Advances in Intelligent Systems and Computing 1253 AISC 362 - 368 2021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）

With the use of information terminals represented by smartphones, our eyes get tired. One index indicating eye fatigue is a change in the number of blinks. In this study, we set the ultimate goal that is to develop eye fatigue prevention system with less time and less physical burden for patient. By doing this, the information terminal performs a flickering of the eyes using a detection camera and a validity check. Previous EAR study with blink detection is confirmed to be effective, and this paper proposes a new formula EARM used EAR. As an evaluation method of blink detection, we used the total time average of the square of the residual value. The results showed that EARM was more accurate than the EAR. Further, it was suggested that the number of blinks during VDT work can be classified into several patterns.

DOI： 10.1007/978-3-030-55307-4_55

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85089624580&origin=inward

Effective speech features for distinguishing mild dementia patients from healthy person 査読有り国際誌

Nishikawa K., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Advances in Intelligent Systems and Computing 1253 AISC 356 - 361 2021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）

The questionnaire method is generally used for present dementia screening. However, this method requires time for 10 to 15 min with a doctor and a clinical psychologist, which puts a burden on hospitals and test subjects. The purpose of this study is to reduce the burden of users by constructing a system to distinguish patients with mild dementia and healthy persons from speech data. Before that this paper examines the effectiveness of speech features. MFCC has been confirmed to be effective in previous research, this paper extracted six kinds of other speech features that are likely to be correlated with symptoms of dementia. This paper got about 90% accuracy rate for a sentence of conversational speech in SVM and Random Forest. Moreover, this paper has calculated the importance of the features by using the SVM-RFE method. As a result, this showed that log-mel spectrum was more important than MFCC.

DOI： 10.1007/978-3-030-55307-4_54

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85089619179&origin=inward

Estimation of Road Damage for Earthquake Evacuation Guidance System Linked with SNS 査読有り国際誌

Yujiro Mihara, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of Congress on Intelligent Systems CIS 2020 2020年09月

Software Log Anomaly Detection Method using HTM Algorithm 査読有り国際誌

Rin Hirakawa, Keitaro Tominaga, Yoshihisa Nakatoh

Proc. of Congress on Intelligent Systems CIS 2020 2020年09月

Study on Software Log Anomaly Detection System with Unsupervised Learning Algorithm 査読有り国際誌

Rin Hirakawa, Keitaro Tominaga, Yoshihisa Nakatoh

Proc. of 2nd International Conference on Human Interaction and Emerging Technologies (IHIET-AI 2020) 2020年08月

Sitting posture discrimination, method for back pain prevention system 査読有り国際誌

Joumori Daiki, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of 2nd International Conference on Human Interaction and Emerging Technologies (IHIET 2020) 2020年08月

Speaker Verification Method Using Hierarchical Temporal Memory for Security System 査読有り国際誌

Yuki Sakaguchi, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of 2nd International Conference on Human Interaction and Emerging Technologies (IHIET 2020) 2020年08月

Blink Detection Using Image Processing to Predict Eye Fatigue 査読有り国際誌

Akihiro Kuwahara, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of 2nd International Conference on Human Interaction and Emerging Technologies (IHIET 2020) 2020年08月

Effective Speech Features for Distinguishing Mild Dementia Patients from Healthy Person 査読有り国際誌

Kazu Nishikawa, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proc. of 2nd International Conference on Human Interaction and Emerging Technologies (IHIET 2020) 2020年08月

Software Log Anomaly Detection Through One Class Clustering of Transformer Encoder Representation 査読有り国際誌

Hirakawa Rin, Tominaga Keitaro, Nakatoh Yoshihisa

Communications in Computer and Information Science ( Springer, Cham ) 1224 ( 1224 ) 655 - 661 2020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）

22nd International Conference on Human-Computer Interaction, HCI International 2020, 19-24 July 2020, Copenhagen, Denmark（新型コロナ感染拡大に伴い、現地開催中止）

その他リンク： https://ci.nii.ac.jp/naid/120007117049

Electrolarynx system using voice conversion based on WaveRNN 査読有り

Urabe E., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2020-January 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）

© 2020 IEEE. Voice is very important in our daily life. However, some people cannot speak because of throat surgery. Electrolarynx (EL) is a way for them to speak again but this voice is flat and noisy like a buzzer. Therefore, we propose a new Electrolarynx system based on WaveRNN. When we use it, the EL voice is converted into a voice close to the natural voice. We assume that this system will be provided as a smartphone application. In the experiment, we evaluated the converted voice by the system and found that naturalness improved.

DOI： 10.1109/ICCE46568.2020.9043135

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85082578662&origin=inward

Study on mistype correction support using attention in Japanese input 査読有り

Komatsu R., Hirakawa R., Kawano H., Nakashi K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2020-January 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）

© 2020 IEEE. How to input characters without mistakes is one of very important points for the interface, when we consider usability of electronic devices such as personal computers and smartphones. Therefore, in this paper, we examine whether we can classify mistypes for Japanese input and inform the users how they make mistakes using 'Attention' and 'LSTM' of machine learning methods. First, we classified the category of mistypes in Japanese input into seven types (Replacement, Removal, Exchange, Insertion, Involvement, Repetition). Second, we asked subjects to input the displayed words and collected their data. Next, we classified the data and labeled it. Finally, we made the Neural Network (LSTM + Attention) to classify the mistype data and inform how they make mistakes. As a result, the correct answer rate of the Neural Network (LSTM + Attention) was about 70% and we were able to know where point the Neural Network focus on by putting in 'Attention'. We are planning to use 'LSTM' to correct mistype directly and use 'Attention' to inform how they frequently make mistakes in the future.

DOI： 10.1109/ICCE46568.2020.9043138

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85082570013&origin=inward

Study on automatic defect report classification system with self attention visualization 査読有り

Hirakawa R., Tominaga K., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2020-January 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）

© 2020 IEEE. In recent years, software in devices such as smartphones and tablets has become increasingly multifunctional, and the use of OSS has become essential. In software development using large-scale OSS, it is important to report defects to appropriate personnel promptly. In this paper, we propose a method to classifying defect reports into appropriate categories using fine-tuned BERT and visualize self-attention information. In the evaluation, category classification was performed using defect reports of the actual OSS project. The F1 score was 0.87, which indicated that high-accuracy classification was possible. Also, the visualization results show that category-specific words can be extracted.

DOI： 10.1109/ICCE46568.2020.9043062

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85082612536&origin=inward

Defect Reports Classification with pre-trained BERT 査読有り国際誌

Rin Hirakawa, Keitaro Tominaga, Yoshihisa Nakatoh

Proceedings of ICISIP2019 2019年09月

Study on the classification of road damage during the earthquake using SNS information 査読有り国際誌

Yujiro Mihara, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proceedings of ICISIP2019 2019年09月

Analyze of Acoustic Features to separate Train Announcement 査読有り国際誌

Yuto Kinoshita, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proceedings of ICISIP2019 2019年09月

Analysis of Words Related to Mistypes 査読有り国際誌

Ryuki Komatsu, Rin Hirakawa, Hideaki Kawano, Kenichi Nakashi, Yoshihisa Nakatoh

Proceedings of ICISIP2019 2019年09月

Study on Door-to-door Sales Detection Method in Watching System for Seniors living alone 査読有り

Hirakawa R., Nakatoh Y.

2019 IEEE International Conference on Consumer Electronics, ICCE 2019 2019年03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）

© 2019 IEEE. In recent years, the number of seniors living alone has increased, so the demand for the watching system has increased. In this paper, we propose a method to detect door-to-door sales from voice information with the aim of constructing a watching system to prevent consumer troubles of seniors living alone. In the proposed method, speech recognition is performed on conversation voice acquired from a microphone, and converted into text data. Subsequently, text data is input to the Doc2Vec model that has learned the conversation data on door-to-door sales and the cosine similarity is calculated. In this system, when the similarity exceeds a certain threshold value, it is determined that door-to-door sales is being conducted and a warning is issued to relatives. In order to evaluate the effectiveness of the proposed method, binary classification experiments of conversational text data were conducted. As a result, the area under the ROC curve drawn based on the cosine similarity was 0.94, suggesting that a valid decision was made.

DOI： 10.1109/ICCE.2019.8662118

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85063793011&origin=inward

Study on Mistype Classification in Japanese Input Using Machine Learning 査読有り国際誌

Komatsu R., Nakatoh Y.

2019 IEEE International Conference on Consumer Electronics, ICCE 2019 2019年03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）

© 2019 IEEE. When considering usability of electronic devices such as computers and smartphones, how to input characters without mistakes is a very important point in the interface. Therefore in this paper, we examine classification of mistypes for Japanese input using machine learning. First, we classified the category of mistypes in Japanese input into six types (replacement, removal, exchange, insertion, involvement, repetition). We asked 12 subjects to input the words displayed and saw a trend. Experimental results showed that there was a common trend such as many replacement errors in all subjects, and characteristic mistypes were found for each subject. Next, we made the neural network (RNN • LSTM) to learn the mistype data of all subjects and classified. As a result, the correct answer rate of RNN was 10%, and the correct answer rate of LSTM was 88%. We are planning to use LSTM for correcting mistype in the future.

DOI： 10.1109/ICCE.2019.8661955

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85063782455&origin=inward

Changes in fundamental frequency and gesture of response corresponding to the understanding level of partner's talk 査読有り国際誌

Natsumi K., Saitoh Y., Wakita Y., Nakatoh Y.

Proceedings - 2018 IEEE International Conference on Artificial Intelligence in Engineering and Technology, IICAIET 2018 83 - 86 2019年02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. This study explores a system that estimates the degree of decline in driving ability in elderly people using nonverbal information from daily conversations. To create this system, it is necessary to find factors that are effective for estimating the degree of decline that has reached a problematic level in daily life. In this paper, we consider the case where elderly people cannot understand the partner's talk as their hearing and concentration abilities drop, and we analyze the relation between the degree of understanding of the partner's talk and the nonverbal characteristic of the response utterance. According to the analysis results, both fundamental frequency (F0) values of response utterance and brightness values of the frame difference image for a response scene when a person can understand their partner's talk tend to be higher than that when he/she does not understand the talk. These results indicate that the F0 value of utterances and brightness values of the frame difference images are important factors to estimate the decline in hearing and concentration abilities.

DOI： 10.1109/IICAIET.2018.8638457

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85063133296&origin=inward

Analysis of Mistyping in Hardware Keyboard 査読有り

Ryuki Komatsu, Yoshihisa Nakatoh

Proceedings of ICISIP2018 2018年09月

Study on Visit Sales Detection Using Similarity of Paragraph Vector 査読有り

Rin Hirakawa, Yoshihisa Nakatoh

Proceedings of ICISIP2018 2018年09月

Study of Volume Restriction Methods Based on the Melody Structure and Vocal of Music on Portable Music Player for Hearing Protection 査読有り

Daiki Okabe, Yoshihisa Nakatoh

Proceedings of ICISIP2018 2018年09月

Study on Wind Noise Reduction using Neural Network for Hearing Aid 査読有り

Katsuya Arimoto, Yoshihisa Nakatoh

Proceedings of ICISIP2018 2018年09月

Comparison of Subjective Evaluation and Adults Voice to Acoustic Features of Children Voice 査読有り

Naoya Harada, Yoshihisa Nakatoh

Proceedings of ICISIP2018 2018年09月

Study of the smartphone operation support system for visually impaired people 査読有り国際誌

Torigoe T., Nakatoh Y.

2018 IEEE International Conference on Consumer Electronics, ICCE 2018 2018-January 1 - 2 2018年03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）

© 2018 IEEE. In this paper, we propose the smartphone operation support system for visually impaired. It is difficult for visually impaired people to grasp complicated screen configuration of smartphone. Visually impaired people have a problem that the people don't know the operation position of touch panel. Therefore, we created the touch screen application for visually impaired people to get the information easily. As a result, when comparing conventional application and proposed application in three environments 'screen display', 'screen non-display', and 'blindfold', we confirmed a drastic reduction in operation time of proposed method.

DOI： 10.1109/ICCE.2018.8326296

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85048738604&origin=inward

Study on headphone hearing loss prevention methods based on the melody structure of music on portable music player 査読有り国際誌

Okabe D., Nakatoh Y.

2018 IEEE International Conference on Consumer Electronics, ICCE 2018 2018-January 1 - 2 2018年03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）

© 2018 IEEE. In this paper, we examine the volume control system in an application to prevent the onset of headphone hearing loss due to music listening using a portable music player. First, the possibility that lower the volume to a safe level without be noticed by the user (difference of perceived attenuating level) are evaluated. As a result, the notice happened when volume level was attenuated 11 ± 3 dB, and it not be affected by changing attenuation rate of volume level. Based on the above experiments, we examine the melody structure of music by attenuate the volume separately for melody structure (Intro part, Verse, Chorus, Interlude). As a result, it was found that the intro part was hard to notice the volume attenuation, and in the part of the chorus, it was easy to notice the volume attenuation. Next, we evaluated difference of the perceived attenuating level considering frequency characteristics. As a result for the low frequency, it was difficult to notice the volume attenuation, and the high frequency tends to notice the volume attenuation.

DOI： 10.1109/ICCE.2018.8326297

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85048777093&origin=inward

Study on Children Voice Features Using Expression Words 査読有り

Harada Naoya, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Comparison of Non-verbal Feature of Free-conversation Speech between Elderly and Young Individuals 査読有り

Yumi Wakita, Masaya Yamasuso, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Study on Program Reading System for Visually Impaired 査読有り

Ryo Hirakawa, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Study of Volume Restriction Methods Based on the Melody Structure of Music for Hearing Protection 査読有り

Daiki Okabe, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Study of the smartphone operation support application for visually impaired 査読有り

Taiki Torigoe, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Analysis of Speech Depending on Age and Speaking Style by Using Acoustic Features 査読有り

Shunsuke Kawahara, Yoshihisa Nakatoh

Proceedings of ICISIP2017 2017年09月

Speech Enhancement of Singing Support System for Lartngectomies 査読有り

Hirotaka Nogamia, Yoshihisa Nakatoh, Kenji Matsui

Proceedings of ICISIP2017 2017年09月

Development of handsfree speech enhancement system for laryngectomies 査読有り

Kenji Matsui, Yuta Matsunaga, Yoshihisa Nakatoh, Yumiko O. Kato

Proceedings of NCSP2017 2017年03月

Study on Evaluation about Unpleasantness of Sounds in Everyday Life by Elderly People 査読有り

Kazuyoshi Sawada, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Improvement of the Approach Vehicle Detection by the Supersonic Wave 査読有り

Kosuke Kuwayama, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Study on Impovement of Sound Localization for Dichotic-Listening 査読有り

Toshiaki Ebihara, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Improvement of MAEQ Method for Speech Enhancement in Noisy Environments 査読有り

Hiroyuki Takamatsu, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Study of Onomatopoeia Expression Caption on TV Contents 査読有り

Kazuki Kodama, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Study of Listening Training to Improve Auditory Ability for Elderly People 査読有り

Shinichiro Asayama, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2016 2016年09月

Study of volume limit for headphone hearing loss prevention on portable music player 査読有り

Higashi T., Tanaka Y., Nakatoh Y.

2016 IEEE International Conference on Consumer Electronics, ICCE 2016 397 - 398 2016年03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 recent years, portable music players become popular due to high usability. However, EU and WHO have pointed out that there is a possibility of causing hearing loss when users listen to music at loud volume and long time. Therefore, it is necessary how to prevent hearing loss. In previous research, it found optimum volume level when listening to music. But it didn't find noticed volume level by listener when volume level of music is gradually attenuated. In this paper, we evaluated whether to be able to lower from loud volume level to safe volume level without the users notice. First, we measured the volume level that the users listen to music daily. We called this volume the optimum listening level. As a result, we found that seven out of the twelve listened to volumes more than 80dB. Next, we measured the difference of level between the 85 dB volume level and the volume level noticed by users when volume level of music is gradually attenuated. We called this difference of level the difference of perceived level by attenuating. As a result, young people noticed when volume level was attenuated 12 ±4 dB without being affected changing attenuation rate of volume level. Elderly people noticed in the range of 13 ± 4 dB as with young people.

DOI： 10.1109/ICCE.2016.7430663

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84965121743&origin=inward

Detecting system of approaching vehicles using an ultrasonic wave 査読有り

Furuno Y., Tanaka Y., Nakatoh Y.

2016 IEEE International Conference on Consumer Electronics, ICCE 2016 393 - 394 2016年03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 elderly population is increasing, and the traffic environment surrounding elderly people is not safe. We can prevent traffic accidents at an early stage by jumping out of the way if we perceive that the approaching vehicle may collide with the walking elderly person. Therefore, we think that it would be useful if we could build a system to inform us of the approaching vehicle by a supersonic wave. In this paper, using a supersonic wave loudspeaker attached to a vehicle, we proposed the technique to detect approaching vehicles on the sidewalk. We found that we could detect areas around 50 m with a high probability of more than 80% by using a tone burst as input to the supersonic wave loudspeaker.

DOI： 10.1109/ICCE.2016.7430661

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84965178007&origin=inward

Improvement of Speech Intelligibility for Elderly People Based on Formant Frequency Shift 査読有り

Yuto Tanaka, Mizumachi Mitshunori,Yoshihisa Nakatoh

Proceedings of NCSP2016 2016年03月

Development of Wearable Speech Enhancement System for Laryngectomees 査読有り

Kenta Kimura, Kenji Matsui, Yoshihisa Nakatoh, Yumiko Ono Kato

Proceedings of NCSP2016 2016年03月

Improvement of Word Intelligibitlity Based on ITD and ILD Control on Binaural Hearing Aids 査読有り

Ryota Fujiwara, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of NCSP2016 2016年03月

Text-reading Support System for the Visually Impaired on Web Sites 査読有り

Takumi Misawa，Yuto Tanaka，Yoshihisa Nakatoh

Proceedings of NCSP2016 2016年03月

Design of elder-friendly auditory signals for microwave ovens 査読有り国際誌

Kadota N., Mizumachi M., Nakatoh Y.

2015 4th International Conference on Informatics, Electronics and Vision, ICIEV 2015 2015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）

© 2015 IEEE. Recent house electrical appliances generate a number of auditory signals, when a user operates an appliance or an appliance draws attention to a user. In this paper, we focus on multi-functional microwave ovens, which commonly have several auditory signals for user's operation, error occurrence, completion of cooking, and so on. A wide variety of auditory signals for current microwave ovens have two problems: it is difficult for users to image what the auditory signals inform, and the auditory signals including high frequency components are not friendly for elderly users. In this paper, popular auditory signals for microwave ovens were subjectively evaluated with/without an elderly-simulated band-pass filter, which simulated an averaged audiogram of elder people, on the five grade scales of easiness of perception, unpleasantness, and emergency. Experimental results suggest that young and elderly people might imagine different function for a single auditory signal. Finally, suitable auditory signals both for young and elderly users are proposed in the viewpoints of frequency and temporal patterns.

DOI： 10.1109/ICIEV.2015.7333977

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84961778212&origin=inward

Study of the Volume Limit for Hearing Protection in the Music Listening 査読有り

Teruhisa Higashi, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2015 2015年09月

Study of Word Intelligibility Based on Control of ITD or ILD on Binaural Hearing 査読有り

Ryota Fujiwara, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2015 2015年09月

Study of Text Reading Method Using Important Words on Web Sites forvisual impaired 査読有り

Takumi Misawa，Yuto Tanaka，Yoshihisa Nakatoh

Proceedings of ICISIP2015 2015年09月

Study of the approach vehicle detection by the supersonic wave 査読有り

Yoshihiro Furuno, Yuto Tanaka, Yoshihisa Nakatoh

Proceedings of ICISIP2015 2015年09月

Singing System by Electric-Larynx Using LPC Residual Wave 査読有り

Takashi Himeno, Yuto Tanaka, Mitsunori Mizumachi, Yoshihisa Nakatoh, Kenji Matsui

Proceedings of ICISIP2015 2015年09月

Study of Improvement of Intelligibility Based on Formant Frequency Shift for Vowels of Elderly Speech 査読有り

Yuto TANAKA, Hiroaki IGAUE, Mitsunori MIZUMACHI, Yoshihisa NAKATOH

Proceedings of Youngnam-Kyushu Joint Conference on Acoustics 2015 157 - 160 2015年01月

Study of Adaptive Beam　Forming with Direction Selector of Target Sound by Hearing Aid User 査読有り

Tsubasa Noguchi, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Conference on Intelligent Systems and Image Processing 2014 155 - 158 2014年09月

Study of Improvement of Intelligibility for Elderly Speech Based on Formant Frequency Shift 査読有り

Yuto Tanaka, Hiroaki Igaue, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Conference on Intelligent Systems and Image Processing 2014 163 - 166 2014年09月

Study of Natural-Voice-Like Vibration Sound for Electrolarynx 査読有り

Takafumi Hei, Yuto Tanaka, Mitsunori Mizumachi, Yoshihisa Nakatoh, Kenji Matsui

International Conference on Intelligent Systems and Image Processing 2014 159 - 162 2014年09月

Study of Word Prediction for Utterance Support System 査読有り

Yuya Hamai, Mitsunori Mizumachi, Yoshihisa Nakatoh, Kenji Matsui

International Conference on Intelligent Systems and Image Processing 2015 57 - 60 2013年09月

Development of Electrolarynx with Hands-Free Prosody Control 査読有り

Kenji Matsui, Kenta Kimura, Yoshihisa Nakatoh, Yumiko O. Kato

8th ISCA Speech Synthesis Workshop 274 - 277 2013年08月

Investigation on Variation of the Initial Phonemes of Japanese Words Due to Aging 査読有り

Yoshiaki Yuda, Ryo Fukuhara, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Conference on Industrial Application Engineering 2013 57 - 60 2013年03月

Preliminary Evaluation of Angry Voice in Automatic Speech Recognition 査読有り

Rikio Ueno, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Conference on Industrial Application Engineering 2013 57 - 60 2013年03月

Noise Reduction by Non-Linear Beamforming Tolerant to Misalignment of Target Direction 査読有り

Hiroaki Hayashi, Mitsunori Mizumachi, Yoshihisa Nakatoh

International Conference on Industrial Application Engineering 2013 57 - 60 2013年03月

A Preliminary User Interface Study of Speech Enhancement System 査読有り

Kenji Matsui, Yuma Otsuki, Yoshihisa Nakatoh, Yumiko O. Kato

International Conference on Industrial Application Engineering 2013 57 - 60 2013年03月

A study of asynchronous phenomenon in articulation for elderly people 査読有り

T. Ide, M. Mizumachi, Y. Nakatoh, K. Niyada

Proceedings of Youngnam-Kyushu Joint Conference on Acoustics 2011 157 - 160 2011年01月

Blind vehicle's sound detec ting technique for advanced safety-driving system 査読有り

Yoshizawa S., Nakatoh Y.

Digest of Technical Papers - IEEE International Conference on Consumer Electronics 2009年09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）

Automobile makers have been developed and commercialized safety-driving systems to reduce the number of traffic accidents, especially with cameras or radars. This paper presents a new vehicle detecting system in a blind spot, which identifies the diffracted sound emitted from the blind vehicles behind obstacle. We show that our system works efficiently in a noisy real environment. ©2009 IEEE.

DOI： 10.1109/ICCE.2009.5012169

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=70349272622&origin=inward

An adaptive MEL-LPC analysis for speech recognition 査読有り

Nakatoh Y., Nishizaki M., Yoshizawa S., Yamada M.

8th International Conference on Spoken Language Processing, ICSLP 2004 933 - 936 2004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）

This paper describes a new speech analysis method, an adaptive Mel-LPC (AMLPC) analysis method, using human auditory characteristics. The Mel-LPC analysis method that we have proposed is an efficient time domain technique to estimate the warped predictors from input speech directly. However, the frequency resolution of spectrum obtained by Mel-LPC analysis is constant regardless of the characteristics of input speech at each analysis frame. In the AMLPC analysis, it is probable to estimate the spectrum coefficients with optimal frequency resolution according to the characteristics of the phoneme at each analysis frame, because the spectral slope or the formant is different according to phoneme (vowels, fricatives and so on). The recognition performance of melcepstrum parameters obtained by the AMLPC analysis was compared with those of mel-cepstrum parameters obtained by the conventional LPC analysis and the Mel-LPC analysis through gender-dependent phoneme and word recognition. The results show that the proposed method leads to a significant improvement in recognition accuracy over conventional LPC analysis, and a slightly improvement of error rate about 10% over the Mel-LPC analysis.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85009107969&origin=inward

A study of vocal tract length normalization with generation-dependent acoustic models 査読有り

Fujita K., Ono Y., Nakatoh Y.

6th International Conference on Spoken Language Processing, ICSLP 2000 2000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）

In this paper, we propose a new speaker adaptation algorithm that employs vocal tract length normalization (VTLN) with generation-dependent acoustic models, and prove its validity with various generation subjects including small children and aged people. Children and aged people have particular features in their pronunciations. For example, children, whose articluatory organs are under growing, often have deficits in articulation. Aged people also have unique pronunciations caused by aging features such as loss of their original teeth. On the other hand, VTL cannot be estimated only by speakers' generation since VTL is highly dependent to speakers' individuality rather than generation. Though children have rather short VTLs than adult and aged people, exact VTL for each speaker cannot be estimated without analysis of each speaker's voice. Based on above our idea on generation features, in this paper, we propose VTLN with generation-dependent acoustic model as a speaker adaptation method suitable for various generations, and discuss the effect of our proposing method. Our results show that proposing method brings word error rate (WER) reduction by 52% for aged people, and by 63% for children.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85009101193&origin=inward

Improvement of intelligibility for elderly speech based on formant frequency shift

Yuto Tanaka, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Utterance Support System using Word Input based on Word Prediction

Yuya Hamai, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Natural-Voice-Like Vibration Sound for Electrolarynx

Takafumi Hei, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Relationship between Utterance Motion and Voice Intelligibility

Hiroaki Igaue, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Adaptive Beamforming with Direction Selector of Target Sound by Hearing Aid User

Tsubasa Noguchi, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Walking Support with Sound for the Visually Impaired

Yuta Shiroishi, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

Word Intelligibility Improvement of Voice Addressing Public in Noisy Environments

Tomohiko Nakamura, Yoshihisa Nakatoh

International Student Computing Project Showcases 2014年09月

高齢者や障害者の支援に向けた音声技術の応用招待有り

中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2013 ( 0 ) 1 - 1 2013年01月

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）

音声認識や音声合成などの音声技術はスマートフォンやインターネットの普及に伴い急速に身近なものになってきた。本講演では，近年急速に増加している高齢者や，声を出して話すことが困難な発声障害者などに向けた音声技術による支援について，その取り組み例を紹介する。また，電子機器に対するアクセシビリティを向上させるための取り組みとして，現在筆者らが推進している音声ユーザインタフェースの国際標準化活動について紹介する。

DOI： 10.12792/iiae2013.001

その他リンク： https://ci.nii.ac.jp/naid/130006694268

Anomaly detection on software log based on Temporal Memory 査読有り

Hirakawa R., Uchida H., Nakano A., Tominaga K., Nakatoh Y.

Computers and Electrical Engineering 95 2021年10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）

For complex system failures, it is necessary to investigate the cause using runtime log data as a clue. The amount of log data output by systems is becoming increasingly large, and there is a need for an automatic log analysis method that can extract only the log data related to the failure. In this study, we propose TM-LAD, a method for automatically detecting log patterns that are different from normal system operation. In our experiments, we compared the performance of anomaly detection with seven existing methods using the loglizer benchmark. The results show that the proposed method has the smallest F-measure of 0.01 standard deviation in the benchmark, and is more robust to changes in the amount of training data than the other methods.

DOI： 10.1016/j.compeleceng.2021.107433

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85114936208&origin=inward

電車内アナウンス音声と周囲の会話を判別するための音響的特徴量の検討

木下雄斗, 中藤良久, 平川凛, 中司賢一, 河野英昭

電気関係学会九州支部連合大会講演論文集 ( 電気・情報関係学会九州支部連合大会委員会 ) 2019 ( 0 ) 257 - 257 2019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）

走行中の電車内では走行音や人の話し声など様々な雑音が存在するため、アナウンス音声がマスキングされることで聞きとりづらくなるという問題がある。本研究では、周囲の話し声を雑音とみなし、電車内アナウンス音声を分離することで聞き取りを改善することを目的として、スピーカから出力されたアナウンス音声と生の音声の間の音響的特徴量の差を解析した。その結果、スペクトル重心・スペクトルロールオフといった音響的特徴量が、スピーカから出力された音声は減少することがわかった。

DOI： 10.11527/jceeek.2019.0_257

その他リンク： https://ci.nii.ac.jp/naid/130007788291

トラック輸送における車体の振動と荷物の傾きの関連性の検討

井下千尋, 平川凛, 中司賢一, 河野英昭, 中藤良久, 中井太地

電気関係学会九州支部連合大会講演論文集 ( 電気・情報関係学会九州支部連合大会委員会 ) 2019 ( 0 ) 273 - 273 2019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）

近年の物流トラックによる事故要因の一例として、「トラックの荷崩れによる落下物事故」が挙げられる。本研究では、その問題解決の為、特にトラック走行時の荷崩れが起こりやすいとされる「傾斜道の乗り越え時」と「急カーブ走行時」を想定し、走行試験を行うことで、車体の振動と荷物の傾きの関連性の検討を行った。実験の測定デバイスとしてドライブレコーダーを採用し、走行時の荷物の映像から荷物の傾きを検出する為に、画像処理における物体追跡手法を用いた。加えて、ドライブレコーダーに内蔵されてある3軸加速度センサで車両の振動を測定し、各々全データ中の1フレームごとの変化量・対応する時間の比較を行った。

DOI： 10.11527/jceeek.2019.0_273

その他リンク： https://ci.nii.ac.jp/naid/130007788220

SNS情報を利用した震災時通路被害の推定に関する検討

三原優二郎, 中藤良久, 中司賢一, 河野英昭, 平川凛

電気関係学会九州支部連合大会講演論文集 ( 電気・情報関係学会九州支部連合大会委員会 ) 2019 ( 0 ) 277 - 277 2019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）

近年、地震によって多くの被害が出ている。それらの被害者の死亡理由のひとつとして、火災や津波といった、二次災害での逃げ遅れが原因として挙げられる。地震によって通路に障害が生じ、通行に時間を要したり、思わぬ危険に見舞われたりする可能性がある。安全に避難するためには、通路の情報を的確に把握し、安全な避難経路を選択することが重要である。また、SNSはリアルタイム性が高く、災害時の情報の取得手段としては非常に有効である。そこで本研究では、SNSの情報をもとに通路の被害内容を把握する方法について検討を行う。

DOI： 10.11527/jceeek.2019.0_277

その他リンク： https://ci.nii.ac.jp/naid/130007788219

自閉症スペクトラムに運動が及ぼす改善効果の定量的評価

河野英昭, 大月匠, 中司賢一, 中藤良久

バイオメディカル・ファジィ・システム学会大会講演論文集 ( バイオメディカル・ファジィ・システム学会 ) 32 ( 0 ) B3 - 3 2019年01月

記述言語：日本語掲載種別：研究論文（その他学術会議資料等）

This study aimed to evaluate the effectiveness of an exercise on the people with autism spectrum disorder using the NeuCube spiking neural network (SNN) architecture. We collected EEG patterns during perception and imitation of facial expressions for each emotion. Comparing the collected data in perceiving and mimicking facial expressions, EEG patterns of typically developed person were very similar. This fact suggest that it seems that there are mirror neurons on facial expression in the human brain. Recently, some studies have been reported that mirror neuron system does not work well in the case of subjects with brain disorders. In this paper, we observed mirror neurons during normal perception and imitation of facial expressions in typically developed subjects (TP), ASD subjects who exercise regularly (ASDe), and ASD subjects who don't exercise regularly (ASDn). As a result of experiments, we confirmed the mirror neuron activities of ASDe subjects tend to be close to ones of TP subjects compared to ones of ASDn subjects. The results in the experiments imply the effects of regular exercise on autism spectrum disorder.

DOI： 10.24466/pacbfsa.32.0_B3-3

その他リンク： https://ci.nii.ac.jp/naid/130007980091

Estimating Age-Dependent Degradation using Nonverbal Feature Analysis of Daily Conversation 査読有り国際誌

脇田由実, 中藤良久

Digital Human Modeling and Applications in Health, Safety, Ergonomics and Risk Management. Healthcare Applications 11582 222 - 231 2019年01月

中国深センへの学生派遣プログラムの取り組み

楊世淵, 中藤良久, 近野泰

教育ブレティン = Educational bulletin ( 九州工業大学 ) 3 - 12 2018年01月

音声スペクトルの時間変化強調による明瞭性改善手法の提案査読有り国際誌

川原竣介, 平川凜, 中藤良久

産業応用工学会論文誌 ( 一般社団法人産業応用工学会 ) 6 ( 1 ) 51 - 54 2018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）

It is hard to listen to an announcement in a noisy environments such as in a train while running. In this paper, to improve speech intelligibility in a noisy environment, we focused on the time variation of speech spectral representing spectral flux from the analysis result which clear speech has larger spectral flux than normal speech. Therefore, we propose speech enhancement method of spectral variation to improve the sound quality evaluation value of STOI. In the evaluation results of STOI score, enhanced speech by proposed method improved more than original speech in all the SNR conditions (SNR = -5, -10, -15 dB). Also, the significant difference in the dangerous rate of 5% was found in SNR = -10, -15 dB.

DOI： 10.12792/jjiiae.6.1.51

その他リンク： http://id.ndl.go.jp/bib/029502703

Comparison of nonverbal feature of free conversation speech between elderly and young individuals 査読有り国際誌

脇田由実, 中藤良久

Proceeding of the 5th IIAE International Conference on Intelligent Systems and Image Processing 2017 5 291 - 296 2017年01月

パワー補正LPC残差を用いた代替発声歌唱法の音質改善

姫野剛至, 田中裕人, 水町光徳, 中藤良久, 松井謙二

電気関係学会九州支部連合大会講演論文集 ( 電気・情報関係学会九州支部連合大会委員会 ) 2015 ( 0 ) 55 - 55 2015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）

我々は、喉頭摘出者の代替発声装置である電気式人工喉頭（EL）を用いた際の歌声の音質改善を行っており、これまでLPC残差を用いた代替発声歌唱法を提案した。本稿では、このLPC残差を用いる代替発声歌唱法において、パワーの補正を行い音質改善を図った。具体的なパワー補正方法として、5母音のうち、比較的パワーの小さい母音/ i /, / u /のパワーを増加させる方法を検討した。評価では、「LPC残差」「パワー補正LPC残差」「Rosenberg波」「ELの歌モード」を用いて歌唱した歌声の比較を行ったところ、パワー補正LPC残差が最も自然性が高いという結果になった。

DOI： 10.11527/jceeek.2015.0_55

その他リンク： https://ci.nii.ac.jp/naid/130006356565

単語発話音声の周波数特徴の加齢変化 (福祉情報工学)

緒方一仁, 水町光徳, 中藤良久

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 ( 261 ) 49 - 54 2014年10月

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）

日本を含む先進諸国では,高齢化が深刻な社会問題となっている.加齢は,音声コミュニケーションにも影響を及ぼす.聴覚の加齢変化については精力的に研究が行われてきたが,発話器官の加齢変化についてはまだ明らかになっていない.本稿では,発話器官の加齢変化に着目し,発話音声の静的特徴について検討する.分析に用いた音声は,64歳から87歳までの男性話者21名による日本語単語発話である.各単語の語頭母音に対して,パワースペクトルを算出し,20代話者と高齢者話者グループの平均と分散を比較した.本分析の結果,加齢により2.5kHzから4.5kHzの周波数帯域でのパワー減少と,4.5kHz以上でのパワー上昇を確認し,すべての音素において周波数特性が平坦になることが判明した.

その他リンク： https://ci.nii.ac.jp/naid/110009958972

A-4-18 フォルマント周波数の補正に基づく高齢者音声の明瞭性改善(A-4.信号処理,一般セッション)

田中裕人, 伊賀上祐彰, 水町光徳, 中藤良久, 松井謙二

電子情報通信学会ソサイエティ大会講演論文集 ( 一般社団法人電子情報通信学会 ) 2014 2014年09月

高齢者音声の音響的特徴に基づいた明瞭性改善方法の検討

伊賀上祐彰, 田中裕人, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 20 - 21 2014年01月

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）

人は加齢に伴って声が変質し,会話の障害になることは少なくない。高齢者音声の先行研究では,高齢者音声の聴感的印象の1つに「めりはり」の無さが指摘されており,声にめりはりがなくなると隣接する音素間のスペクトル差である「遷移量」が低下することがわかっている。また,遷移量と第1,第2フォルマントには相関があることが分かっている。そこで,本研究ではフォルマント周波数に着目し,補正することで,高齢者音声の明瞭性改善の検討を行う。結果として,めりはりのない高齢者音声の音響的特徴として,/a/,/i/,/e/の第1,第2フォルマントが/u/に収束していることがわかった。また、/a/,/i/,/e/を含む口や舌を大きく動かす単語について、第1,第2フォルマントを補正することによって,明瞭性を改善することができた。

DOI： 10.12792/iiae2014.011

その他リンク： https://ci.nii.ac.jp/naid/130006694295

高齢者支援のための車両接近音の音響的特徴の解析

古野慎弥, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 22 - 23 2014年01月

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、高齢者が歩行時に自主的に交通事故を防止できるシステムを構築することを最終目標とし、本稿ではその基礎検討として車両の検出に接近車両の音が使えるかどうか検討を行った。車両接近音は1kHz付近の周波数帯域部分にエネルギーが集中しており、そこに着目すれば30m付近の接近車両を検出できる可能性があることがわかった。今後は、風の強い日や雨天など異なる環境下での比較や、誤検出の問題を解決するための方法を検討する予定である。

DOI： 10.12792/iiae2014.012

その他リンク： https://ci.nii.ac.jp/naid/130006694321

電気式人工喉頭を用いた歌声の音質改善

姫野剛至, 屏貴文, 田中裕人, 水町光徳, 中藤良久, 松井謙二

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 16 - 17 2014年01月

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、喉頭摘出者のための発声器具である電気式人工喉頭(以下EL)を用いた歌声の改善を行う。具体的な方法として、ELに使用される振動体の替わりに振動スピーカを使用し、歌声にはどのような声帯音源が適しているかを検討し、歌声の音質改善を図る。「LPC残差波」「Rosenberg波」「EL」の3つの音源を用いて歌唱した歌声について、聴取実験を行ったところ、LPC残差波による歌が最も自然性が高いことが分かった。また、音響的特徴について生歌とLPC残差波による歌を比較すると、/i/、/u/の音圧やスペクトルについて異なっていることが分かった。今後は、より自然性の高い歌声の実現のためにLPC残差波の音圧やスペクトルの補正方法について検討する。

DOI： 10.12792/iiae2014.009

その他リンク： https://ci.nii.ac.jp/naid/130006694288

視覚障碍者のための重要単語抽出に基づくWebページ読み上げ方法の検討

三澤匠, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 18 - 19 2014年01月

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）

一般的に視覚障碍者はWebから情報を得るときに読み上げソフトを使用しているが，従来の読み上げ方法では欲しい情報を得るまでに長い時間を要する。そこで，情報収集の効率化を図るためWebサイトの中にある重要なキーワードを短時間でより理解できる音声再生方法を提案し，どれだけ聞き取ることができるか比較検討を行った。結果，左右の耳でそれぞれ別の情報を与えることで，短時間に情報を詰め込むことができ，単語を聞き取ることが可能であると分かった。

DOI： 10.12792/iiae2014.010

その他リンク： https://ci.nii.ac.jp/naid/130006694289

方向情報を付与した補助音声による音声明瞭性の改善

藤原良太, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 14 - 15 2014年01月

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）

聴力が低下すると、音の方向感がうまく掴めなくなり、聞きたい音とその他の音を聞き分けることが困難になる。また、補聴器を装用した場合においても十分な改善は見られないことが分かっている。本研究では、難聴者の聞き分けの支援を目的として、方向情報を付与した補助音声をイントラコンカ型イヤホンで再生することによって、目的音の明瞭性が改善するのか聴取実験により検討した。その結果、擬似的に方向情報を付与した補助音声を再生することで、目的音の明瞭性が改善されることが分かった。

DOI： 10.12792/iiae2014.008

その他リンク： https://ci.nii.ac.jp/naid/130006694287

文字入力を用いた発話支援システム実使用時の評価

濱井裕也, 水町光徳, 中藤良久, 松井謙二

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 38 - 39 2014年01月

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）

本稿では、構音障害により発話が困難な人のための予測変換と文字入力を用いた発話支援システムの実使用時の評価を行った。その結果、作成したプロトタイプを用いることで、一部の被験者について単位時間発話文字数が従来法に比べて向上した。今後は、より早い発話を実現するため、音声合成の処理時間やGUIの視認性を改善、対話エンジンを用いた会話制御方法の導入等を行う予定である。

DOI： 10.12792/iiae2014.020

その他リンク： https://ci.nii.ac.jp/naid/130006694297

携帯音楽プレーヤーにおける聴力保護のための音量制限の検討

東照久, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2014 ( 0 ) 24 - 25 2014年01月

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）

近年、携帯音楽プレーヤーは音楽をどこにでも持ち運べるという利便性から普及が進んでいる。しかし一方で、過剰な音量での長時間におよぶ音楽聴取が聴力損失を招く危険性があると指摘されている。本研究では、聴取者が日頃どれくらいの音量で音楽を聞いているのか把握するために、最適聴取レベルの測定と音量を徐々に減衰させた時、どの程度の速度で音量を下げると被験者が知覚できるのかについて評価を行った。実験の結果、騒音下での平均最適聴取レベルは80dBを超えていることが分かった。また、単位時間あたりの減衰量(減衰速度)の違いによる反応時間を測定したところ、減衰速度によらず、7dB±3dBの範囲で被験者が知覚することが分かった。

DOI： 10.12792/iiae2014.013

その他リンク： https://ci.nii.ac.jp/naid/130006694322

K-064 発話支援のための予測変換を用いた文字入力方法の検討(K分野:教育工学・福祉工学・マルチメディア応用,一般論文)

濱井裕也, 水町光徳, 中藤良久, 松井謙二

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 ) 701 - 704 2013年08月

A-19-12 発話支援のための予測変換を用いた文字入力方法の予備検討(A-19.福祉情報工学)

濱井裕也, 中藤良久

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 ) 2013 2013年03月

D-14-11 モーションセンサを用いた電気式人工喉頭の試作(D-14.音声)

松井謙子, 大槻勇真, 中藤良久, 加藤弓子

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 ) 2013 ( 1 ) 2013年03月

電気喉頭音声の音質改善に関する予備検討

田中裕人, 屏貴文, 水町光徳, 中藤良久

電気関係学会九州支部連合大会講演論文集 ( 電気・情報関係学会九州支部連合大会委員会 ) 2013 ( 0 ) 14 - 14 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）

喉頭摘出者は電気式人工喉頭（以下電気喉頭）を用いて発声しているが、電気喉頭の直接音が雑音となるという問題と、音声が不明瞭であるという問題がある。　本研究では、電気喉頭音声を拡声する際に、体のどの部分で収録したら良いか検討した。耳孔内マイク、骨伝導マイク、咽喉マイクを用いて収録した結果、耳孔内マイクで収録した電気喉頭音声が最もSNRが改善されることが分かった。一方、電気喉頭に使用される振動体に替えて、様々な音源を振動スピーカで再生する方法を用いた場合の音質比較を行った。音源として方形波、ノコギリ波、理想的な声帯振動とされるRosenberg波などを用いた結果、Rosenberg波の音が、低域において改善されることが分かった。

DOI： 10.11527/jceeek.2013.0_14

その他リンク： https://ci.nii.ac.jp/naid/130005480247

音源波形の違いが電気喉頭音声に与える影響の検討

屏貴文, 田中裕人, 水町光徳, 中藤良久, 松井謙二

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2013 ( 0 ) 24 - 25 2013年01月

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）

近年，喉頭摘出者のための電気式人工喉頭（以下電気喉頭）による音声の音質改善の検討が行われており，中村らにより肉伝導人工音声によるコミュニケーション支援システムが提案されている。現在我々は，電気喉頭を用いて発声された音声をマイクで収音し，スピーカで拡声する発声補助システムを検討している。本稿では，電気喉頭音声よりも音質の良い音源の検討を行った。ケプストラム分析により声道特性を比較するとRosenberg波音源が低域でも通常の声帯振動に近い音源であることがわかった。また，ケプストラムのケフレンシー幅に着目し，声帯特性を比較するとRosenberg波の声帯特性が最も通常発声のケフレンシー幅に近く，約96%の幅であった。以上のことより，電気喉頭の音源にRosenberg波を用いることで，電気喉頭音声の音質の改善が行える可能性があることがわかった。

DOI： 10.12792/iiae2013.014

その他リンク： https://ci.nii.ac.jp/naid/130006694275

日本人姓を含む音声のワードスポッティング音声認識法の検討

田畑竜二, 水町光徳, 中藤良久

産業応用工学会全国大会講演論文集 ( 一般社団法人産業応用工学会 ) 2013 ( 0 ) 22 - 23 2013年01月

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）

近年，日本の急速な高齢化に伴う老人性難聴者の増加は大きな社会問題の一つである。難聴者に対する支援の一つとして補聴器の使用があるが，高度難聴者や加齢による認知能力の低下した高齢者の会話が聞き取れない原因は聴力の低下だけではないため，その効果は十分とは言えない。また，難聴者が病院等の受付で名前を呼ばれた時に気付くことができないため，不便に感じているというニーズがある。そこで，補聴器のような小型機器にユーザの名前が呼ばれた際に反応するための音声認識エンジンを組み込み，さらに聴力の低下したユーザのために視覚情報や振動により名前が呼ばれたことを知らせてくれる新たな認知支援システムがあれば有用と考える。そこで本論文では，日本人姓を含む音声のワードスポッティング音声認識法の性能改善を目的として，日本人姓の特徴に基づいた2種類の改善方法 (敬称に基づいた改善方法，無音区間の自動検出に基づいた改善方法)を提案し，その評価を行った。その結果，認識性能の低下の大きな原因の一つであった湧き出しを抑え，誤認識を防ぐことができた。

DOI： 10.12792/iiae2013.013

その他リンク： https://ci.nii.ac.jp/naid/130006694274

D-14-11 高齢者の音響的特徴と音声認識性能との関係性の検討(D-14.音声,一般セッション)

井手隆宏, 水町光徳, 中藤良久

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 ) 2012 ( 1 ) 2012年03月

D-14-10 遅い声に着目した語頭における高齢化の影響の解析(D-14.音声,一般セッション)

湯田吉亮, 水町光徳, 中藤良久

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 ) 2012 ( 1 ) 2012年03月

共分散共有型HMMにおける計算量削減方法の検討

山田麻紀, 石田明, 中藤良久

日本音響学会研究発表会講演論文集 2005 ( 1 ) 135 - 136 2005年03月

応答文の切り替えがもたらすタスク達成時間短縮効果についての検討

井上剛, 小沼知浩, 西崎誠, 中藤良久

日本音響学会研究発表会講演論文集 2002 ( 1 ) 215 - 216 2002年03月

対数パワースペクトル領域での減算処理による乗法性歪みの除去

中藤良久, 藤田景子

日本音響学会研究発表会講演論文集 2002 ( 1 ) 183 - 184 2002年03月

音素対ごとに特化した特徴量を用いた音素認識法の検討

芳澤伸一, 脇田由実, 山田麻紀, 中藤良久

日本音響学会研究発表会講演論文集 2000 ( 2 ) 61 - 62 2000年09月

スペクトルに応じて周波数分解能を変化させる線形予測分析法の検討

中藤良久, 芳澤伸一, 山田麻紀

日本音響学会研究発表会講演論文集 2000 ( 2 ) 51 - 52 2000年09月

周波数伸縮を用いた話者正規化法の種々の年齢層話者に対する検討

大野剛男, 藤田景子, 中藤良久

日本音響学会研究発表会講演論文集 2000 ( 1 ) 105 - 106 2000年03月

メルLPCケプストラムの耐雑音性の評価

中藤良久, 松本弘

日本音響学会研究発表会講演論文集 1999 ( 1 ) 23 - 24 1999年03月

オーディオ符号化におけるMDCT係数の削減

劉亜興, 中藤良久, 松本弘

日本音響学会研究発表会講演論文集 1999 ( 1 ) 219 - 220 1999年03月

音声認識におけるメルLPCケプストラムの評価

降旗佳範, 中藤良久, 松本弘

日本音響学会研究発表会講演論文集 1998 ( 2 ) 5 - 6 1998年09月

CV, VC音響モデルの分布共有化による音声認識時の演算量削減

片山大朗, 中藤良久, 芹川光彦, 中橋順一, 八木順子

日本音響学会研究発表会講演論文集 1998 ( 2 ) 83 - 84 1998年09月

音声認識におけるメル線形予測分析法の評価

中藤良久, 松本弘

電子情報通信学会技術研究報告. SP, 音声 ( 一般社団法人電子情報通信学会 ) 98 ( 105 ) 65 - 72 1998年06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）

Strubeによって提案されたバークあるいはメルスケール上の線形予測分析は、人間の聴覚特性を用いたMFCCやPLP分析と同様に音声認識に有効であると期待される。しかし、この方法は通常の線形予測分析と比較して計算量が多く、音声認識にほとんど使用されていない。本報告では、メル化された予測係数を時間領域で効率的に推定するための分析手法(メル線形予測分析:MLPC分析)を提案する。本手法は、通常の線形予測分析の約2倍の計算量で実現できる。MLPC分析により得られるメルケプストラム(MLPCケプストラム)と通常のLPCメルケプストラムとを男性不特定話者音声認識実験を通して比較した。その結果、MLPCケプストラムの方はLPCメルケプストラムよりも音素認識率で約8.6%高い73.4%の認識率を、また単語認識率(520単語)で約3.3%高い96.0%の認識率を示した。

その他リンク： https://ci.nii.ac.jp/naid/110003296933

オーディオ符号化におけるスペクトル包絡推定法の改良

中藤良久, 則松武志, 劉亜興, 松本弘

日本音響学会研究発表会講演論文集 1998 ( 1 ) 245 - 246 1998年03月

AN EFFICIENT MEL-LPC ANALYSIS METHOD FOR SPEECH RECOGNITION 査読有り国際誌

Matsumoto H., Nakatoh Y., Furuhata Y.

5th International Conference on Spoken Language Processing, ICSLP 1998 1998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 proposes a simple and efficient time domain technique to estimate an all-poll model on a mel-frequency axis (Mel-LPC). This method requires only two-fold computational cost as compared to conventional linear prediction analysis. The recognition performance of mel-cepstral parameters obtained by the Mel LPC analysis is compared with those of conventional LP mel-cepstra and the mel-frequency cepstrum coefficients (MFCC) through gender-dependent phoneme and word recognition tests. The results show that the Mel-LPC cepstrum attains a significant improvement in recognition accuracy over conventional LP mel-cepstrum, and gives slightly higher accuracy for male speakers and slightly lower accuracy for female speakers than MFCC.

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85128348808&origin=inward

スペクトル変換を用いたささやき音声の高明瞭化の検討

中藤良久

日本音響学会研究発表会講演論文集 1997 ( 1 ) 301 - 302 1997年03月

線形写像による帯域制限音声の疑似広帯域化

中藤良久, 津島峰生, 則松武志

電子情報通信学会技術研究報告. EA, 応用音響 95 ( 480 ) 1 - 8 1996年01月

帯域制限された音声における広帯域化法の比較

中藤良久, 津島峰生, 則松武志

日本音響学会研究発表会講演論文集 1995 ( 2 ) 355 - 356 1995年09月

帯域拡大による広帯域音声の合成方法の検討

津島峰生, 中藤良久, 則松武志

日本音響学会研究発表会講演論文集 1995 ( 1 ) 249 - 250 1995年03月

音韻認識における連続分布型HMMの教師なし話者適応 (応用音響に関する特集)

中藤良久, 松本弘

信州大学工学部紀要 ( 信州大学工学部 ) ( 75 ) p77 - 86 1995年02月

TV会議用話者自動検出装置の開発

則松武志, 中藤良久, 北島克美

電子情報通信学会技術研究報告. US, 超音波 ( 一般社団法人電子情報通信学会 ) 93 ( 440 ) 25 - 31 1994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）

TV会議システムにおいて、話者の音声を検出して自動的に画面を切り替える高精度の話者自動検出装置を開発した。誤動作のない画面切り替えを実現するために、まず母音と無声摩擦音を検出し、それぞれの音韻の存在比率からファジィ推論を用いて会話音声と雑音とを判別する高精度の音声, 雑音判別方式を提案した。これにより会議中に予想される様々な雑音の除去率98.3%を達成した。また、隣接するマイクロホン間の相互相関から音の到達時間差を計算し、話者の位置を推定することにより、隣接話者間の誤検出を防止することが可能となった