Faculty Profiles - INOUE Sozo

Papers - INOUE Sozo

Division display21 - 40 of about 238／ All the affair displays >>

Estimation of Record Contents for Automatic Generation of Care Records Reviewed International journal

Haru Kaneko, Tahera Hossain, Sozo Inoue

Activity and Behavior Computing, Springer Nature 289 - 306 2021.01

Language：English Publishing type：Research paper (scientific journal)

Elderly person are increasing all over the world. Smart nursing facilities
are required to support them for their good health.Especially in Japan, it is essential to
have automated nursing facilities center to support the relentlessly increasing elderly
person all over the country.As long as elderly people are increasing, the demand for
nursing care services are also increasing day by day.There is a shortage of nurse
to support this vast elderly people.In order to resolve the shortage of nurse, it is
important to reduce the work load for nurse.There are some research to simplify the
care recording process.The main approach is activity recognition or the development
of new apps. However, we aimed for automatic generation of the care records of near
future. So in this paper, we evaluate the accuracy of the automatic generation of
care records using care record data from real nursing facility to simplify the task of
making care records.We made two machine learning models to estimate the "target
patient" and the "recorded value" of care records.And we evaluated the classification
result from the data collected by the care record app.We used two months of data we
collected from Japanese nursing facility for the evaluation.We can achieve, average
F1-score of 'target patient' estimation is 74% and the average F1-score of 'detail of
record' estimation is 58%. We believe that using these results in care records app
will simplify the task of making care records.

Other Link： https://link.springer.com/chapter/10.1007/978-981-15-8944-7_18

Exploring Human Activities Using eSense Earable Device Reviewed International journal

Md. Shafiqul Islam, Tahera Hossain, Md Atiqur Rahman Ahad, Sozo Inoue

Activity and Behavior Computing, Springer Nature 169 - 185 2021.01

Improvement of Human Action Recognition Using 3D Pose Estimation Reviewed International journal

Kohei Adachi, Paula Lago, Tsuyoshi Okita, Sozo Inoue

Activity and Behavior Computing, Springer Nature 21 - 37 2021.01

Summary of the Cooking Activity Recognition Challenge Reviewed International journal

Sayeda Shamma Alia, Paula Lago, Shingo Takeda, Kohei Adachi, Brahim Brenaissa, Md Atiqur Rahman Ahad, Sozo Inoue

Human Activity Recognition Challenge, Springer Nature 1 - 13 2021.01

Multilabel Classification of Nursing Activities in a Realistic Scenario Reviewed International journal

Farina Faiz, Yoshinori Ideno, Sozo Inoue

Activity and Behavior Computing, Springer Nature 269 - 288 2021.01

On-Device Deep Personalization for Robust Activity Data Collection Reviewed International journal

Nattaya Mairittha, Tittaya Mairittha, Sozo Inoue

Sensors, MDPI 21 ( 41 ) 2 - 2 2020.12

Language：English Publishing type：Research paper (scientific journal)

One of the biggest challenges of activity data collection is the need to rely on users and keep them engaged to continually provide labels. Recent breakthroughs in mobile platforms have proven effective in bringing deep neural networks powered intelligence into mobile devices. This study proposes a novel on-device personalization for data labeling for an activity recognition system using mobile sensing. The key idea behind this system is that estimated activities personalized for a specific individual user can be used as feedback to motivate user contribution and improve data labeling quality. First, we exploited fine-tuning using a Deep Recurrent Neural Network to address the lack of sufficient training data and minimize the need for training deep learning on mobile devices from scratch. Second, we utilized a model pruning technique to reduce the computation cost of on-device personalization without affecting the accuracy. Finally, we built a robust activity data labeling system by integrating the two techniques outlined above, allowing the mobile application to create a personalized experience for the user. To demonstrate the proposed model's capability and feasibility, we developed and deployed the proposed system to realistic settings. For our experimental setup, we gathered more than 16,800 activity windows from 12 activity classes using smartphone sensors. We empirically evaluated the proposed quality by comparing it with a baseline using machine learning. Our results indicate that the proposed system effectively improved activity accuracy recognition for individual users and reduced cost and latency for inference for mobile devices. Based on our findings, we highlight critical and promising future research directions regarding the design of efficient activity data collection with on-device personalization.

Kyutacar

Other Link： https://doi.org/10.3390/s21010041

Implementation of Care Records Automatic Generation Function in a Care Record Application Reviewed International journal

Haru Kaneko, Tahera Hossain, Sozo Inoue

The 40th Joint Conference on Asia-Pacific Association of Medical Informatics (APAMI) 6 pages 2020.12

Language：English Publishing type：Research paper (international conference proceedings)

Online Online

The number of elderly people is increasing in the world. Especially in Japan, the shortage of nurses has been a problem in recent years. So far, we had developed a mobile application to make nursing care records. In addition, we are also experimenting with the introduction of this application to nursing homes. In this paper, we design an automatic generation function of the nursing care record into our mobile nursing care record application. In this paper, we implement an automatic generation function of the recorded contents in this care record system. we aim to simplify the work of making care records. To do so, we make a machine learning model for estimating the feature nursing care records in the server. And, we design new UI on our application to make care records using estimation result. As a result of the evaluation, we know it is possible to reduce time that making care records using our application with automatic generation function of care records.

Activity Recognition for Assisting People with Dementia Reviewed International journal

Muhammad Fikry, Defry Hamdhana, Paula Lago, Sozo Inoue

Contactless Human Activity Analysis, Springer Nature 0 2 - 2 2020.12

Language：English Publishing type：Research paper (scientific journal)

Online Online

Technology can help and provide support to people with dementia to ensure their safety during daily
activities. In this paper, we summarize information about activity recognition for people living with
dementia. This paper aims to understand the uses and types of applications, the types of sensors/systems,
methods, and data used within the scope of human activity recognition to monitor, detect symptoms, or
help people with dementia. To this end, 447 abstracts were collected from a Scopus database, which
yielded 127 relevant papers and 102 papers that were considered in detail based on the four categories of
assessment (application, system/sensors, methods, and data). This paper shows the trend that smart
environment technology is most widely used for monitoring people with dementia, wherein machine
learning techniques as a method for activity recognition to achieve the results of testing or implementing
the system. We conclude that combining sensor devices and the addition of smartphone devices in one
system is suitable for implementation because it can be used as an identity such that it distinguishes the
object under study with other objects. During the monitoring process, prevention can be achieved
simultaneously by adding a warning alarm in the smartphone when people with dementia perform
abnormal activities, and the results need to be further analyzed to get the best pattern of activities for
people with dementia. Next, the type of application that was initially designed for monitoring can be
developed into an assistant for people with dementia.

Enhancing Nursing Care Records with A Spoken Dialogue System based on Smartphones Reviewed International journal

Tittaya Mairittha, Nattaya Mairittha, Sozo Inoue

The 40th Joint Conference on Asia-Pacific Association of Medical Informatics (APAMI) 8 pages 2020.11

Language：English Publishing type：Research paper (international conference proceedings)

Background and Objective: This study describes the integration of a spoken dialogue system and
nursing records on an Android smartphone application intending to help nurses reduce documentation
time and improve the overall experience of a healthcare setting. The application also incorporates with
collecting personal sensor data and activity labels for activity recognition.
Methods: We developed a joint model based on a bidirectional long-short term memory and conditional random fields (Bi-LSTM-CRF) to identify user intention and extract record details from user
utterances. Then, we transformed unstructured data into record inputs on the smartphone application.
Results: The joint model achieved the highest F1-score at 96.79%. Moreover, we conducted an experiment to demonstrate the proposed model's capability and feasibility in recording in realistic settings.
Our preliminary evaluation results indicate that when using the dialogue-based, we could increase the
percentage of documentation speed to 58.13% compared to the traditional keyboard-based.
Conclusions: Based on our findings, we highlight critical and promising future research directions
regarding the design of the efficient spoken dialogue system and nursing records.

Mobile application for caregiver in collecting statistical data of BPSD attack focused on macro activities: PhD forum abstract Reviewed International journal

Defry Hamdhana

Proceedings of the Conference on Embedded Networked Sensor Systems (SenSys) 807 - 808 2020.11

Language：English Publishing type：Research paper (international conference proceedings)

Japan Yokohama (virtual)

BPSD affects Dementia patients often and varies. It is needed a best and suit treatment from doctors to handle BPSD attack toward Dementia patients. Therefore, to obtain accurate data of patients in daily care, caregivers are expected to conduct ...

Other Link： https://doi.org/10.1145/3384419.3430577

Multilabel classification in human activity recognition: PhD forum abstract Reviewed International journal

Farina Faiz

Proceedings of the Conference on Embedded Networked Sensor Systems (SenSys) 809 - 810 2020.11

Language：English Publishing type：Research paper (international conference proceedings)

Japan Yokohama (virtual)

Human activities become very difficult to handle when they appear without proper temporal information. A smarter approach is to handle this problem is using multilabel classification. However, most of the multilabel methods are suitable for text data ...

Other Link： https://doi.org/10.1145/3384419.3430578

Requirements analysis for reminder system in daily activity recognition dementia: PhD forum abstract Reviewed International journal

Muhammad Fikry

Proceedings of the Conference on Embedded Networked Sensor Systems (SenSys) 815 - 816 2020.11

Language：English Publishing type：Research paper (international conference proceedings)

Japan Yokohama (virtual)

Technology assistance refers to all tools or systems that allow a people with dementia to carry out their daily activities in order to increase convenience and security in these activities. Dementia is a condition of a collection of symptoms of irreversible brain damage that occurs over a relatively long time period, resulting in problems concentrating, planning or organizing daily activities. This paper is focused on estimating daily activity recognition time by considering the feasibility of activity types, feature extraction techniques for speech recognition, and the design effectiveness of a system that is easy to use by people with dementia. The results from activity recognition need to be further analyzed to find the best pattern for people with dementia. After that, monitoring applications can develop into prevention of abnormal activity in dementia.

Other Link： https://doi.org/10.1145/3384419.3430572

Improving Activity Data Collection with On-device Personalization Using Fine-tuning Reviewed International journal

Nattaya Mairittha, Tittaya Mairittha, Sozo Inoue

Ubicomp Workshop on Human Activity Sensing Corpus and Applications (HASCA) 255 - 260 2020.09

WorkerSense: Mobile Sensing Platform for Collecting Physiological, Mental, and Environmental State of Office Workers Reviewed International journal

Yuri Tani, Shuichi Fukuda, Yuki Matsuda, Sozo Inoue

PerHealth 2020: 5th IEEE PerCom Workshop on Pervasive Health Technologies (PerHealth 2020) 2020.09

Language：English Publishing type：Research paper (international conference proceedings)

In data collection of the human physiological and psychological conditions
for mental healthcare (e.g., work engagement), measurement methods using
environment-installed sensors and questionnaire surveys have been often
used. However, these approaches are not practical in continuous data
collection, due to the large burden for people. Recently, in association
with advancing sensing technology with IoTs, sensing by small sensors and
wearable devices has become possible easily. In this paper, we
aim to establish a simple and general sensing method based on a mobile
application for measuring physiological and psychological state of office
workers and environmental state. Through the experiment for 2-3 weeks
involving 60 office workers of four Japanese companies by using our
application, we succeeded to create a dataset of physiological,
environment, and mental state. This paper explains the developed mobile
application, experimental procedure, and a summary of the data collected in
the experiment.

Improving Fine-tuned Question Answering Models for Electronic Health Records Reviewed International journal

Tittaya Mairittha, Nattaya Mairittha, Sozo Inoue

Ubicomp Workshop on Computing for Well-Being (WellComp) 688 - 691 2020.09

Language：English Publishing type：Research paper (international conference proceedings)

Mexico (virtual) Mexico (virtual)

The prevalence of voice assistants has strengthened the interest in a
question answering for the medical domain, allowing both patients
and healthcare providers to enter a question naturally and pinpoint
useful information quickly. However, a large number of medical
terms makes the creation of such a system a demanding task. To
address this challenge, we explore transfer learning techniques
for constructing a personalized EHR-QA system. The goal is to
answer questions regarding a discharge summary in an electronic
health record (EHR). We present the experiments with a pre-trained
BERT (Bidirectional Encoder Representations from Transformers)
model fine-tuned on different tasks and show the results obtained
to provide insights into learning effects and training effectiveness.

Kyutacar

Other Link： https://doi.org/10.1145/3410530.3414436

MCoMat: a New Performance Metric for Imbalanced Multi-layer Activity Recognition Dataset Reviewed International journal

Sayeda Shamma Alia, Paula Lago, Sozo Inoue

Ubicomp Workshop on Human Activity Sensing Corpus and Applications (HASCA) 232 - 237 2020.09

Language：English Publishing type：Research paper (international conference proceedings)

Mexico (virtual) Mexico (virtual)

Existing performance metrics assess classifiers on single granularity layer. Having multi-layer labels is also possible such as activity recognition datasets. Semantic annotations could be given with multiple granularity layers in these datasets e.g., ...

Kyutacar

Other Link： https://doi.org/10.1145/3410530.3414364

Summary of the 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data Reviewed International journal

Sayeda Shamma Alia, Paula Lago, Kohei Adachi, Tahera Hossain, Hiroki Goto, Tsuyoshi Okita, Sozo Inoue

Ubicomp Workshop on Human Activity Sensing Corpus and Applications (HASCA) 378 - 383 2020.09

Language：English Publishing type：Research paper (international conference proceedings)

Mexico (virtual) Mexico (virtual)

2nd Nurse Care Activity Recognition Challenge using Lab and Field Data is organized as a part of HASCA workshop and is continuation of Nurse Care Activity Recognition Challenge [7]. We give the description of the dataset and summarize the approaches ...

Kyutacar

Other Link： https://doi.org/10.1145/3410530.3414611

Automatic Labeled Dialogue Generation for Nursing Record Systems Reviewed International journal

Tittaya Mairittha, Nattaya Mairittha, Sozo Inoue

JPM Journal of Personalized Medicine 10 ( 3 ) 6 - 2 2020.09

Language：English Publishing type：Research paper (scientific journal)

The integration of digital voice assistants in nursing residences is becoming increasingly important to facilitate nursing productivity with documentation. A key idea behind this system is training natural language understanding (NLU) modules that enable the machine to classify the purpose of the user utterance (intent) and extract pieces of valuable information present in the utterance (entity). One of the main obstacles when creating robust NLU is the lack of sufficient labeled data, which generally relies on human labeling. This process is cost-intensive and time-consuming, particularly in the high-level nursing care domain, which requires abstract knowledge. In this paper, we propose an automatic dialogue labeling framework of NLU tasks, specifically for nursing record systems. First, we apply data augmentation techniques to create a collection of variant sample utterances. The individual evaluation result strongly shows a stratification rate, with regard to both fluency and accuracy in utterances. We also investigate the possibility of applying deep generative models for our augmented dataset. The preliminary character-based model based on long short-term memory (LSTM) obtains an accuracy of 90% and generates various reasonable texts with BLEU scores of 0.76. Secondly, we introduce an idea for intent and entity labeling by using feature embeddings and semantic similarity-based clustering. We also empirically evaluate different embedding methods for learning good representations that are most suitable to use with our data and clustering tasks. Experimental results show that fastText embeddings produce strong performances both for intent labeling and on entity labeling, which achieves an accuracy level of 0.79 and 0.78 f1-scores and 0.67 and 0.61 silhouette scores, respectively.

Kyutacar

Other Link： https://doi.org/10.3390/jpm10030062

Towards Mapping Activity Classes for Transfer Learning in Human Activity Recognition Reviewed International journal

Md. Shafiqul Islam, Sayeda Shamma Alia, Sozo Inoue

International Conference on Activity and Behavior Computing (ABC) 5 pages 2020.08

Synthesizing Cell Protein data for Human Protein Cell Profiling Using Dual Deep Generative Modeling Reviewed International journal

Rakesh Ranjan, Sozo Inoue, Tom Shibata

International Conference on Imaging, Vision & Pattern Recognition (IVPR) 1 - 6 2020.08

Language：English Publishing type：Research paper (international conference proceedings)

Japan Kitakyushu

To understand the biology of health, and how molecular dysfunction leads to disease, knowledge of the human cell is essential. The protein is the core unit of the human body made from trillions of cells, forming the body's various tissues. These tissues come together to create human organs. It is essential to understand the Spatio-temporal distribution of proteins in cells and to investigate human RNA-sequencing for human genes characterization. For this, it requires a massive amount of annotated data. However, due to many considerations like the high cost of data sample collection, lack of data sample availability, and lawful clauses for patient privacy, the majority of medical data is out of reach for general public research. In this study, we propose a new dual deep generative method for synthesizing human cell protein images by using the Generative Adversarial Network technique. Specifically, for that, we pair original cell protein images with their respective Cell-protein-tree. These pairs are then used to learn the mapping from a binary cell protein to a new cell protein image. For this purpose, we use an image-to-image translation technique based on adversarial learning. The generated cell protein images are expected to preserve the structural and visual quality of the training images. Visual and quantitative analysis of the experimental results demonstrates that the synthesized data are preserving the desired quality while maintaining the different forms of original data. Contribution-We have proposed a new dual deep generative model for synthesizing cell protein data.

Other Link： https://ieeexplore.ieee.org/document/9306574

　 PREV - NEXT 　

<INOUE Sozo>

Papers - INOUE Sozo