Yiyuan Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yiyuan Zhang has a total of 58 co-authors. Here are the top co-authors by number of shared publications:
- Wei Chen
The University of Sydney
11 shared publications - Chen Chen
Human Phenome Institute, Fudan University, Shanghai 201203, China
9 shared publications - Bart Vanrumste
Department of Electrical Engineering (ESAT), KU Leuven, 3000 Leuven, Belgium
7 shared publications - Cong Fu
Department of Neurology, Huashan Hospital, Fudan University, Shanghai, China
2 shared publications - Dominique Schreurs
KU Leuven
2 shared publications - Hongyu Chen
Greater Bay Area Institute of Precision Medicine, Guangzhou, China
2 shared publications
Latest Publications
- 8 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (12) , 8743-8750
MtRBD: Advancing iRBD Analysis With Multi-Task Learning for Joint Sleep Staging and RSWA Detection
Rapid eye movement (REM) sleep without atonia (RSWA) is a critical diagnostic criterion for REM sleep behavior disorder (RBD). Current clinical practices rely on time-consuming manual annotation, increasing workload, and introducing variability. Existing automated methods, including CNN, LSTM, Transformer, and their combinations, fail to fully exploit the inherent physiological relationship between sleep staging and RSWA detection, while also facing challenges such as severe data imbalance and the complex fusion of multichannel signals. To address these limitations, we propose a multi-task learning framework that jointly optimizes both tasks. Our Multi-scale Information Attention Bottleneck Network (MIABNet) backbone improves multichannel fusion. Building upon MIABNet, the Multi-task RBD Network (MtRBD) implements dynamic feature enhancement, facilitating cross-task information flow while preserving physiological relationships. On the clinical CZ-RBD dataset collected from 30 patients over 59 nights at Shanghai Changzheng Hospital, our framework achieved the accuracy of 83.0% for sleep staging and 93.6% for RSWA detection, with an accuracy of 77.6% for critical RSWA event identification, outperforming single-task methods in cross-subject validation. Through modality selection and Grad-CAM visualization, we enhance clinical interpretability, providing reliable support for early detection of RBD and related neurodegenerative diseases.
- 8 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (12) , 8743-8750
MtRBD: Advancing iRBD Analysis With Multi-Task Learning for Joint Sleep Staging and RSWA Detection
Rapid eye movement (REM) sleep without atonia (RSWA) is a critical diagnostic criterion for REM sleep behavior disorder (RBD). Current clinical practices rely on time-consuming manual annotation, increasing workload, and introducing variability. Existing automated methods, including CNN, LSTM, Transformer, and their combinations, fail to fully exploit the inherent physiological relationship between sleep staging and RSWA detection, while also facing challenges such as severe data imbalance and the complex fusion of multichannel signals. To address these limitations, we propose a multi-task learning framework that jointly optimizes both tasks. Our Multi-scale Information Attention Bottleneck Network (MIABNet) backbone improves multichannel fusion. Building upon MIABNet, the Multi-task RBD Network (MtRBD) implements dynamic feature enhancement, facilitating cross-task information flow while preserving physiological relationships. On the clinical CZ-RBD dataset collected from 30 patients over 59 nights at Shanghai Changzheng Hospital, our framework achieved the accuracy of 83.0% for sleep staging and 93.6% for RSWA detection, with an accuracy of 77.6% for critical RSWA event identification, outperforming single-task methods in cross-subject validation. Through modality selection and Grad-CAM visualization, we enhance clinical interpretability, providing reliable support for early detection of RBD and related neurodegenerative diseases.
- 10 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (22) , 42050-42062
An Eye Mask-Based Wearable System Toward Tailored Sleep Monitoring
Comfortable and reliable wearable systems for home-based automatic sleep staging are crucial for remotely diagnosing sleep disorders. Among various categories of biosignals, electrooculography (EOG) has emerged as a promising modality due to its ease of acquisition and robustness to noise. However, research on EOG-based sleep staging remains limited, with most studies constrained to the standard 2-channel EOG configuration recommended by the American Academy of Sleep Medicine (AASM) scoring manual. Thus, it may lack higher spatial information, hindering fully capturing the tiny sleep-related eye movements. In this study, we developed a novel eye mask-based EOG acquisition system capable of collecting 14-channel spatially diverse EOG signals. Building upon this hardware innovation, we further proposed a dedicated deep learning model, MTNet, combining multi-scale feature extraction and temporal attention, to fully exploit the rich information from the collected EOG data for accurate sleep staging , which is capable of effectively capturing relevant sleep features from EOG signals. We also evaluated the importance of each EOG channel for sleep staging, providing recommendations for EOG-based data acquisition in-home sleep monitoring systems. Our study involved 24 participants, resulting in 48 overnight sleep recordings. In K-fold cross-validation, 14-channel EOG achieved an average accuracy of 83.2%, a Cohen’s kappa coefficient of 0.77, and an MF1 score of 77.7%. These results surpass the performance of the standard EOG, which yielded an accuracy of 82.3%, a Cohen’s kappa coefficient of 0.76, and an MF1 score of 76.1%. In particular, the 14-channel EOG demonstrated superior detection accuracy in the N1 and REM sleep stages compared to the standard setup. In conclusion, the proposed 14-channel EOG collection system—formed as a wearable eye mask—shows significant potential as a high-dimensional biosignal source for home-based sleep monitoring. By combining 14-channel EOG acquisition with the MTNet model, and investigating the importance of channels, this study marks a substantial step forward in an accurate and user-friendly sleep monitoring solution.
- 10 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (22) , 42050-42062
An Eye Mask-Based Wearable System Toward Tailored Sleep Monitoring
Comfortable and reliable wearable systems for home-based automatic sleep staging are crucial for remotely diagnosing sleep disorders. Among various categories of biosignals, electrooculography (EOG) has emerged as a promising modality due to its ease of acquisition and robustness to noise. However, research on EOG-based sleep staging remains limited, with most studies constrained to the standard 2-channel EOG configuration recommended by the American Academy of Sleep Medicine (AASM) scoring manual. Thus, it may lack higher spatial information, hindering fully capturing the tiny sleep-related eye movements. In this study, we developed a novel eye mask-based EOG acquisition system capable of collecting 14-channel spatially diverse EOG signals. Building upon this hardware innovation, we further proposed a dedicated deep learning model, MTNet, combining multi-scale feature extraction and temporal attention, to fully exploit the rich information from the collected EOG data for accurate sleep staging , which is capable of effectively capturing relevant sleep features from EOG signals. We also evaluated the importance of each EOG channel for sleep staging, providing recommendations for EOG-based data acquisition in-home sleep monitoring systems. Our study involved 24 participants, resulting in 48 overnight sleep recordings. In K-fold cross-validation, 14-channel EOG achieved an average accuracy of 83.2%, a Cohen’s kappa coefficient of 0.77, and an MF1 score of 77.7%. These results surpass the performance of the standard EOG, which yielded an accuracy of 82.3%, a Cohen’s kappa coefficient of 0.76, and an MF1 score of 76.1%. In particular, the 14-channel EOG demonstrated superior detection accuracy in the N1 and REM sleep stages compared to the standard setup. In conclusion, the proposed 14-channel EOG collection system—formed as a wearable eye mask—shows significant potential as a high-dimensional biosignal source for home-based sleep monitoring. By combining 14-channel EOG acquisition with the MTNet model, and investigating the importance of channels, this study marks a substantial step forward in an accurate and user-friendly sleep monitoring solution.
- Open Access
- 26 May 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (6) , 571
AttenCRF-U: Joint Detection of Sleep-Disordered Breathing and Leg Movements in OSA Patients
Obstructive sleep apnea (OSA) is characterized by frequent episodes of sleep-disordered breathing (SDB), which are often accompanied by leg movement (LM) events, especially periodic limb movements during sleep (PLMS). Traditional single-event detection methods often overlook the dynamic interactions between SDB and LM, failing to capture their temporal overlap and differences in duration. To address this, we propose Attention-enhanced CRF with U-Net (AttenCRF-U), a novel joint detection framework that integrates multi-head self-attention (MHSA) within an encoder–decoder architecture to model long-range dependencies between overlapping events and employs multi-scale convolutional encoding to extract discriminative features across different temporal scales. The model further incorporates a conditional random field (CRF) to refine event boundaries and enhance temporal continuity. Evaluated on clinical PSG recordings from 125 OSA patients, the model with CRF improved the average F1 score from 0.782 to 0.788 and reduced temporal alignment errors compared with CRF-free baselines. The joint detection strategy distinguished respiratory-related leg movements (RRLMs) from PLMS, boosting the PLMS detection F1 score from 0.756 to 0.778 and the SDB detection F1 score from 0.709 to 0.728. By integrating MHSA into a CRF-augmented U-Net framework and enabling joint detection of multiple event types, this study presents a novel approach to modeling temporal dependencies and event co-occurrence patterns in sleep disorder diagnosis.
- 26 May 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (6) , 571
Open AccessAttenCRF-U: Joint Detection of Sleep-Disordered Breathing and Leg Movements in OSA Patients
Obstructive sleep apnea (OSA) is characterized by frequent episodes of sleep-disordered breathing (SDB), which are often accompanied by leg movement (LM) events, especially periodic limb movements during sleep (PLMS). Traditional single-event detection methods often overlook the dynamic interactions between SDB and LM, failing to capture their temporal overlap and differences in duration. To address this, we propose Attention-enhanced CRF with U-Net (AttenCRF-U), a novel joint detection framework that integrates multi-head self-attention (MHSA) within an encoder–decoder architecture to model long-range dependencies between overlapping events and employs multi-scale convolutional encoding to extract discriminative features across different temporal scales. The model further incorporates a conditional random field (CRF) to refine event boundaries and enhance temporal continuity. Evaluated on clinical PSG recordings from 125 OSA patients, the model with CRF improved the average F1 score from 0.782 to 0.788 and reduced temporal alignment errors compared with CRF-free baselines. The joint detection strategy distinguished respiratory-related leg movements (RRLMs) from PLMS, boosting the PLMS detection F1 score from 0.756 to 0.778 and the SDB detection F1 score from 0.709 to 0.728. By integrating MHSA into a CRF-augmented U-Net framework and enabling joint detection of multiple event types, this study presents a novel approach to modeling temporal dependencies and event co-occurrence patterns in sleep disorder diagnosis.
- 6 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-14
ASTGSleep: Attention-Based Spatial–Temporal Graph Network for Sleep Staging
Accurate sleep staging for patients with sleep disorders requires the integration of multiple channels and various signal modalities. While Graph Convolutional Networks (GCNs) have significantly improved sleep staging accuracy by effectively processing complex data, current GCN-based methods often require a large number of signal channels and struggle to capture temporal correlations effectively. They are leading to a large limitation in the scope of graph network models in sleep staging. To address these challenges, we propose an attention based Spatial-Temporal Graph Network for sleep staging, named ASTGSleep. ASTGSleep effectively addresses the temporal continuity of sleep data and overcomes limitations related to the number of channels by integrating the Temporal Graph Construction and Feature Extraction Module (TGFEM). TGFEM captures dynamic temporal features through temporal graph construction, complementing spatial features and enhancing sensitivity to sleep stage transitions. Moreover, the Multi-Dimensional Fusion Module (MDFM) optimizes and fuses these spatio-temporal features using a graph attention network, further improving accuracy and robustness in sleep staging. Experimental results demonstrate that ASTGSleep achieves competitive accuracy on datasets with different numbers of channels, such as MASSSS3 with 26 channels (87.0%) and SleepEDF-20 with 3 channels (86.5%). Furthermore, its performance on sleep disorder datasets, including CAP (81.4%) and clinical CZ-RBD (82.7%), shows robust generalization and enhanced accuracy, making it a promising tool for automatic sleep staging in clinical diagnostics.
- 6 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-14
ASTGSleep: Attention-Based Spatial–Temporal Graph Network for Sleep Staging
Accurate sleep staging for patients with sleep disorders requires the integration of multiple channels and various signal modalities. While Graph Convolutional Networks (GCNs) have significantly improved sleep staging accuracy by effectively processing complex data, current GCN-based methods often require a large number of signal channels and struggle to capture temporal correlations effectively. They are leading to a large limitation in the scope of graph network models in sleep staging. To address these challenges, we propose an attention based Spatial-Temporal Graph Network for sleep staging, named ASTGSleep. ASTGSleep effectively addresses the temporal continuity of sleep data and overcomes limitations related to the number of channels by integrating the Temporal Graph Construction and Feature Extraction Module (TGFEM). TGFEM captures dynamic temporal features through temporal graph construction, complementing spatial features and enhancing sensitivity to sleep stage transitions. Moreover, the Multi-Dimensional Fusion Module (MDFM) optimizes and fuses these spatio-temporal features using a graph attention network, further improving accuracy and robustness in sleep staging. Experimental results demonstrate that ASTGSleep achieves competitive accuracy on datasets with different numbers of channels, such as MASSSS3 with 26 channels (87.0%) and SleepEDF-20 with 3 channels (86.5%). Furthermore, its performance on sleep disorder datasets, including CAP (81.4%) and clinical CZ-RBD (82.7%), shows robust generalization and enhanced accuracy, making it a promising tool for automatic sleep staging in clinical diagnostics.
- 4 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (8) , 13554-13565
SASE-TCN: A Multitask Learning Network for MDS-UPDRS III Task Classification and Parkinson’s Disease Recognition
As wearable technology and machine learning (ML) algorithms have developed by leaps and bounds, many studies have focused on automatically monitoring Parkinson’s Disease (PD). The procedure includes two steps: firstly, classifying activities, usually the Movement Disorder Society-sponsored revision of the unified Parkinson’s disease rating scale (MDS-UPDRS) part III tasks; secondly, evaluating the performance of these activities and recognizing PD. To improve the efficiency and precision of the procedure, this study proposes a self-attention squeeze-and-excitation temporal convolutional network (SASE-TCN). Based on the multi-stage temporal convolutional network (MS-TCN) and attention mechanisms, SASE-TCN can effectively extract the distant sequential and channel-wise features. In addition, based on acceleration signals, SASE-TCN can classify MDS-UPDRS III tasks and recognize patients with PD simultaneously with the multitask learning mechanism. The model was tested with two public datasets, PD-BioStampRC21 and PD-motion. The SASE-TCN model obtained an average F1 score of 0.7802-0.9160 for activity classification and an average accuracy of 0.7366-0.8578 for PD recognition. The results demonstrated the feasibility of the SASE-TCN in classifying PD-related activities and recognizing PD with accelerometers. This study will support the diagnosis and treatment of PD.
- 4 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (8) , 13554-13565
SASE-TCN: A Multitask Learning Network for MDS-UPDRS III Task Classification and Parkinson’s Disease Recognition
As wearable technology and machine learning (ML) algorithms have developed by leaps and bounds, many studies have focused on automatically monitoring Parkinson’s Disease (PD). The procedure includes two steps: firstly, classifying activities, usually the Movement Disorder Society-sponsored revision of the unified Parkinson’s disease rating scale (MDS-UPDRS) part III tasks; secondly, evaluating the performance of these activities and recognizing PD. To improve the efficiency and precision of the procedure, this study proposes a self-attention squeeze-and-excitation temporal convolutional network (SASE-TCN). Based on the multi-stage temporal convolutional network (MS-TCN) and attention mechanisms, SASE-TCN can effectively extract the distant sequential and channel-wise features. In addition, based on acceleration signals, SASE-TCN can classify MDS-UPDRS III tasks and recognize patients with PD simultaneously with the multitask learning mechanism. The model was tested with two public datasets, PD-BioStampRC21 and PD-motion. The SASE-TCN model obtained an average F1 score of 0.7802-0.9160 for activity classification and an average accuracy of 0.7366-0.8578 for PD recognition. The results demonstrated the feasibility of the SASE-TCN in classifying PD-related activities and recognizing PD with accelerometers. This study will support the diagnosis and treatment of PD.
- Open Access
- 30 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 32 (15344320) , 3410-3421
Lightweight Neural Network for Sleep Posture Classification Using Pressure Sensing Mat at Various Sensor Densities
Recently, pressure-sensing mats have been widely used to capture static and dynamic pressure over sleep for posture recognition. Both a full-size mat with a low-density sensing array for figuring out the structure of the whole body and a miniature scale mat with a high-density sensing array for identifying the local characteristics around the chest have been investigated. However, both of the mat systems may face the challenge in the trade-off between the computational complexity (involving the size, density, etc. of the mat) and the performance of sleep posture recognition, where high performance may requires overcomplex computation and result in time latency in real-time sleep posture monitoring. In this paper, a lightweight neural network named ConcatNet, is proposed to realize sleep postures (supine, left, right, and prone) recognition in real time while maintaining a favorable performance. In ConcatNet, the inception module is proposed to extract the image features under multiple receptive fields, while the multi-layer feature fusion module is utilized to fuse deep and shallow features to enhance the model performance. To further improve the efficiency of the model, the depthwise convolution is adopoted. ConcatNet models in 3 different scales (ConcatNet-S, ConcatNet-M, and ConcatNet-L) are built to explore the impact of the sensor density on sleep posture recognition performance. Experimental results exhibit that ConcatNet-M corresponding to medium sensor density (16×16) achieved the best comprehensive performance, with short-term data cross-validation accuracy at 95.56% and overnight data testing accuracy at 94.68%. The model size is 7.91KB, FLOPs is 56.47K, and the inference time is only 0.38ms, which shows an outstanding performance of real-time sleep posture recognition with minimum consumption, indicating the potential to be deployed in mobile devices.
- 30 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 32 (15344320) , 3410-3421
Open AccessLightweight Neural Network for Sleep Posture Classification Using Pressure Sensing Mat at Various Sensor Densities
Recently, pressure-sensing mats have been widely used to capture static and dynamic pressure over sleep for posture recognition. Both a full-size mat with a low-density sensing array for figuring out the structure of the whole body and a miniature scale mat with a high-density sensing array for identifying the local characteristics around the chest have been investigated. However, both of the mat systems may face the challenge in the trade-off between the computational complexity (involving the size, density, etc. of the mat) and the performance of sleep posture recognition, where high performance may requires overcomplex computation and result in time latency in real-time sleep posture monitoring. In this paper, a lightweight neural network named ConcatNet, is proposed to realize sleep postures (supine, left, right, and prone) recognition in real time while maintaining a favorable performance. In ConcatNet, the inception module is proposed to extract the image features under multiple receptive fields, while the multi-layer feature fusion module is utilized to fuse deep and shallow features to enhance the model performance. To further improve the efficiency of the model, the depthwise convolution is adopoted. ConcatNet models in 3 different scales (ConcatNet-S, ConcatNet-M, and ConcatNet-L) are built to explore the impact of the sensor density on sleep posture recognition performance. Experimental results exhibit that ConcatNet-M corresponding to medium sensor density (16×16) achieved the best comprehensive performance, with short-term data cross-validation accuracy at 95.56% and overnight data testing accuracy at 94.68%. The model size is 7.91KB, FLOPs is 56.47K, and the inference time is only 0.38ms, which shows an outstanding performance of real-time sleep posture recognition with minimum consumption, indicating the potential to be deployed in mobile devices.
- 21 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28 (9) , 5189-5200
PSEENet: A Pseudo-Siamese Neural Network Incorporating Electroencephalography and Electrooculography Characteristics for Heterogeneous Sleep Staging
Sleep staging plays a critical role in evaluating the quality of sleep. Currently, most studies are either suffering from dramatic performance drops when coping with varying input modalities or unable to handle heterogeneous signals. To handle heterogeneous signals and guarantee favorable sleep staging performance when a single modality is available, a pseudo-siamese neural network (PSN) to incorporate electroencephalography (EEG), electrooculography (EOG) characteristics is proposed (PSEENet). PSEENet consists of two parts, spatial mapping modules (SMMs) and a weight-shared classifier. SMMs are used to extract high-dimensional features. Meanwhile, joint linkages among multi-modalities are provided by quantifying the similarity of features. Finally, with the cooperation of heterogeneous characteristics, associations within various sleep stages can be established by the classifier. The evaluation of the model is validated on two public datasets, namely, Montreal Archive of Sleep Studies (MASS) and SleepEDFX, and one clinical dataset from Huashan Hospital of Fudan University (HSFU). Experimental results show that the model can handle heterogeneous signals, provide superior results under multimodal signals and show good performance with single modality. PSEENet obtains accuracy of 79.1%, 82.1% with EEG, EEG and EOG on Sleep-EDFX, and significantly improves the accuracy with EOG from 73.7% to 76% by introducing similarity information.
- 21 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 28 (9) , 5189-5200
PSEENet: A Pseudo-Siamese Neural Network Incorporating Electroencephalography and Electrooculography Characteristics for Heterogeneous Sleep Staging
Sleep staging plays a critical role in evaluating the quality of sleep. Currently, most studies are either suffering from dramatic performance drops when coping with varying input modalities or unable to handle heterogeneous signals. To handle heterogeneous signals and guarantee favorable sleep staging performance when a single modality is available, a pseudo-siamese neural network (PSN) to incorporate electroencephalography (EEG), electrooculography (EOG) characteristics is proposed (PSEENet). PSEENet consists of two parts, spatial mapping modules (SMMs) and a weight-shared classifier. SMMs are used to extract high-dimensional features. Meanwhile, joint linkages among multi-modalities are provided by quantifying the similarity of features. Finally, with the cooperation of heterogeneous characteristics, associations within various sleep stages can be established by the classifier. The evaluation of the model is validated on two public datasets, namely, Montreal Archive of Sleep Studies (MASS) and SleepEDFX, and one clinical dataset from Huashan Hospital of Fudan University (HSFU). Experimental results show that the model can handle heterogeneous signals, provide superior results under multimodal signals and show good performance with single modality. PSEENet obtains accuracy of 79.1%, 82.1% with EEG, EEG and EOG on Sleep-EDFX, and significantly improves the accuracy with EOG from 73.7% to 76% by introducing similarity information.
- 11 November 2023
- Published by Elsevier in Expert Systems with Applications
- Vol. 240
MAGSleepNet: Adaptively multi-scale temporal focused sleep staging model for multi-age groups
- 11 November 2023
- Published by Elsevier in Expert Systems with Applications
- Vol. 240
MAGSleepNet: Adaptively multi-scale temporal focused sleep staging model for multi-age groups
- Open Access
- 27 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 30 (15344320) , 1776-1785
Can Wearable Devices and Machine Learning Techniques Be Used for Recognizing and Segmenting Modified Physical Performance Test Items?
Assessment of physical performance is essential to predict the frailty level of older adults. The modified Physical Performance Test (mPPT) clinically assesses the performance of nine activities: standing balance, chair rising up & down, lifting a book, putting on and taking off a jacket, picking up a coin, turning 360°, walking, going upstairs, and going downstairs. The activity performing duration is the primary evaluation standard. In this study, wearable devices are leveraged to recognize and predict mPPT items’ duration automatically. This potentially allows frequent follow up of physical performance, and facilitates more appropriate interventions. Five devices, including accelerometers and gyroscopes, were attached to the waist, wrists and ankles of eight younger adults. The system was experimented within three aspects: machine learning models, sensor placement, and sampling frequencies, to which the non-causal six-stages temporal convolutional network using 6.25 Hz signals from the left wrist and right ankle obtained the optimal performance. The duration prediction error ranged from 0.63 ± 0.29 s (turning 360°) to 8.21 ± 16.41 s (walking). The results suggest the potential for the proposed system in the automatic recognition and segmentation of mPPT items. Future work includes improving the recognition performance of lifting a book and implementing the frailty score prediction.
- 27 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Vol. 30 (15344320) , 1776-1785
Open AccessCan Wearable Devices and Machine Learning Techniques Be Used for Recognizing and Segmenting Modified Physical Performance Test Items?
Assessment of physical performance is essential to predict the frailty level of older adults. The modified Physical Performance Test (mPPT) clinically assesses the performance of nine activities: standing balance, chair rising up & down, lifting a book, putting on and taking off a jacket, picking up a coin, turning 360°, walking, going upstairs, and going downstairs. The activity performing duration is the primary evaluation standard. In this study, wearable devices are leveraged to recognize and predict mPPT items’ duration automatically. This potentially allows frequent follow up of physical performance, and facilitates more appropriate interventions. Five devices, including accelerometers and gyroscopes, were attached to the waist, wrists and ankles of eight younger adults. The system was experimented within three aspects: machine learning models, sensor placement, and sampling frequencies, to which the non-causal six-stages temporal convolutional network using 6.25 Hz signals from the left wrist and right ankle obtained the optimal performance. The duration prediction error ranged from 0.63 ± 0.29 s (turning 360°) to 8.21 ± 16.41 s (walking). The results suggest the potential for the proposed system in the automatic recognition and segmentation of mPPT items. Future work includes improving the recognition performance of lifting a book and implementing the frailty score prediction.
- Open Access
- 27 November 2021
- Published by Institution of Engineering and Technology (IET) in Healthcare Technology Letters
- Vol. 8 (6) , 148-158
Detecting hand washing activity among activities of daily living and classification of WHO hand washing techniques using wearable devices and machine learning algorithms
During COVID‐19, awareness of proper hand washing has increased significantly. It is critical that people learn the correct hand washing techniques and adopt good hand washing habits. Hence, this study proposes using wearable devices to detect hand washing activity among other daily living activities (ADLs) and classify steps proposed by the World Health Organization (WHO). Two experiments were conducted with 16 participants, aged from 20 to 31. The first experiment was hand washing following WHO regulation (ten participants), and the second experiment was performing eight ADLs (eight participants). All participants wore two wearable devices equipped with accelerometers and gyroscopes; one on each wrist. Four machine learning classifiers were compared in classifying hand washing steps in the leave‐one‐subject‐out (LOSO) mode. The SVM model with Gaussian kernel achieved the best performance in classifying 11 washing hands steps, with an average F1‐score of 0.8501. When detected among the other ADLs, hand washing following WHO regulation obtained the F1‐score of 0.9871. The study demonstrates that wearable devices are feasible to detect hand washing activity and the hand washing techniques as well. The classification results of getting the soap and rubbing thumbs are low, which will be the main focus in the future study.
- 27 November 2021
- Published by Institution of Engineering and Technology (IET) in Healthcare Technology Letters
- Vol. 8 (6) , 148-158
Open AccessDetecting hand washing activity among activities of daily living and classification of WHO hand washing techniques using wearable devices and machine learning algorithms
During COVID‐19, awareness of proper hand washing has increased significantly. It is critical that people learn the correct hand washing techniques and adopt good hand washing habits. Hence, this study proposes using wearable devices to detect hand washing activity among other daily living activities (ADLs) and classify steps proposed by the World Health Organization (WHO). Two experiments were conducted with 16 participants, aged from 20 to 31. The first experiment was hand washing following WHO regulation (ten participants), and the second experiment was performing eight ADLs (eight participants). All participants wore two wearable devices equipped with accelerometers and gyroscopes; one on each wrist. Four machine learning classifiers were compared in classifying hand washing steps in the leave‐one‐subject‐out (LOSO) mode. The SVM model with Gaussian kernel achieved the best performance in classifying 11 washing hands steps, with an average F1‐score of 0.8501. When detected among the other ADLs, hand washing following WHO regulation obtained the F1‐score of 0.9871. The study demonstrates that wearable devices are feasible to detect hand washing activity and the hand washing techniques as well. The classification results of getting the soap and rubbing thumbs are low, which will be the main focus in the future study.