Xiaodong Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiaodong Liu has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Fang Qiu
Shanghai University of International Business and Economics
2 shared publications- Xinming Ye
East China University of Science and Technology
2 shared publications Ce Wang
Sport and Health Research Center, Physical Education Department, Tongji University, Shanghai, China
1 shared publication- Chen Chen
State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
1 shared publication - Chenyun Dai
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, P. R. China
1 shared publication - Fang Qiu
School of Physical Education & Health, Shanghai University of International Business and Economics, Shanghai 201620, China
1 shared publication
Latest Publications
- 1 August 2026
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 23 (4) , 046061
Knowledge distillation for sEMG-based gesture recognition: enhancing wearable HMI systems with lightweight models
Objective . Gesture recognition using surface electromyography (sEMG) is a promising technology for wearable human–machine interaction systems. However, sEMG-based gesture classification models often suffer from performance degradation in cross-user scenarios due to the variability in individual physiological signals. To address this issue, large-scale deep learning algorithms with a great amount of parameters are employed to enhance model generalization and improve recognition performance across different users. However, the use of deep learning models on wearable devices is challenging due to their limited computational power and memory capacity. Methods . This study addresses this challenge through knowledge distillation, which compresses large teacher models into lightweight student models while maintaining performance. We trained deep teacher models, including DenseNet, InceptionV3, and VggNet, and distilled their parameters into compact convolutional neural network and long short-term memory (LSTM) models. The experiment was performed on the sEMG data of 30 subjects collected from a wrist-band electrode with 32 channels. Main Results . The results demonstrated that student models, particularly LSTM-based ones, achieved classification accuracy close to or even higher than their teacher models, with a highest accuracy improvement. Among the teacher–student combinations, the DenseNet121-LSTM architecture achieved the highest classification accuracy. The relationship between computational complexity floating-point operations (FLOPs) and model performance was also analyzed, showing that the distilled models can effectively approximate high-FLOP models. Significance . This research enhances the feasibility of embedding deep learning models into wearable systems (e.g. wristband or wristwatch), facilitating more responsive and efficient gesture recognition in daily-life applications. The Python implementation of the complete knowledge distillation framework is publicly available at: https://github.com/Open-EXG/handDistill .
- 1 August 2026
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 23 (4) , 046061
Knowledge distillation for sEMG-based gesture recognition: enhancing wearable HMI systems with lightweight models
Objective . Gesture recognition using surface electromyography (sEMG) is a promising technology for wearable human–machine interaction systems. However, sEMG-based gesture classification models often suffer from performance degradation in cross-user scenarios due to the variability in individual physiological signals. To address this issue, large-scale deep learning algorithms with a great amount of parameters are employed to enhance model generalization and improve recognition performance across different users. However, the use of deep learning models on wearable devices is challenging due to their limited computational power and memory capacity. Methods . This study addresses this challenge through knowledge distillation, which compresses large teacher models into lightweight student models while maintaining performance. We trained deep teacher models, including DenseNet, InceptionV3, and VggNet, and distilled their parameters into compact convolutional neural network and long short-term memory (LSTM) models. The experiment was performed on the sEMG data of 30 subjects collected from a wrist-band electrode with 32 channels. Main Results . The results demonstrated that student models, particularly LSTM-based ones, achieved classification accuracy close to or even higher than their teacher models, with a highest accuracy improvement. Among the teacher–student combinations, the DenseNet121-LSTM architecture achieved the highest classification accuracy. The relationship between computational complexity floating-point operations (FLOPs) and model performance was also analyzed, showing that the distilled models can effectively approximate high-FLOP models. Significance . This research enhances the feasibility of embedding deep learning models into wearable systems (e.g. wristband or wristwatch), facilitating more responsive and efficient gesture recognition in daily-life applications. The Python implementation of the complete knowledge distillation framework is publicly available at: https://github.com/Open-EXG/handDistill .
- Open Access
- 29 December 2025
- Published by Frontiers Media SA in Frontiers in Neuroscience
- Vol. 19
Influence of electrode placement on the recognition of different gesture categories using high-density sEMG
High-density surface electromyography (HD-sEMG)-based gesture recognition serves as a critical interface for human-computer interaction (HCI). However, recognition accuracy exhibits a significant dependency on gesture complexity and electrode positioning. To address this, we systematically investigated the relationship between gesture types and sEMG electrode placement locations through intra-subject, inter-day, and inter-subject validation protocols. Two distinct gesture categories were analyzed, i.e., single-degree-of-freedom (single-DoF) gestures and daily-used multi-finger synergistic gestures. Using an open-access gesture dataset, HD-sEMG signals were acquired from three forearm regions: the distal wrist, mid-forearm, and proximal elbow, separately. Classification results using support vector machine (SVM) revealed that single-DoF gestures achieved peak accuracy with distal wrist signals (98.63% for intra-subject, 79.73% for inter-day, and 75.47% for inter-subject validation protocols), whereas daily-used gestures performed optimally with signals from the mid-forearm and proximal elbow regions. These findings demonstrate the specific relationship between electrode placement and gesture type, providing valuable insights for EMG-HCI design and sensor placement strategies based on the nature of the target gesture.
- 29 December 2025
- Published by Frontiers Media SA in Frontiers in Neuroscience
- Vol. 19
Open AccessInfluence of electrode placement on the recognition of different gesture categories using high-density sEMG
High-density surface electromyography (HD-sEMG)-based gesture recognition serves as a critical interface for human-computer interaction (HCI). However, recognition accuracy exhibits a significant dependency on gesture complexity and electrode positioning. To address this, we systematically investigated the relationship between gesture types and sEMG electrode placement locations through intra-subject, inter-day, and inter-subject validation protocols. Two distinct gesture categories were analyzed, i.e., single-degree-of-freedom (single-DoF) gestures and daily-used multi-finger synergistic gestures. Using an open-access gesture dataset, HD-sEMG signals were acquired from three forearm regions: the distal wrist, mid-forearm, and proximal elbow, separately. Classification results using support vector machine (SVM) revealed that single-DoF gestures achieved peak accuracy with distal wrist signals (98.63% for intra-subject, 79.73% for inter-day, and 75.47% for inter-subject validation protocols), whereas daily-used gestures performed optimally with signals from the mid-forearm and proximal elbow regions. These findings demonstrate the specific relationship between electrode placement and gesture type, providing valuable insights for EMG-HCI design and sensor placement strategies based on the nature of the target gesture.
- Open Access
- 12 August 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (8) , 869
Effects of Aging on Motor Unit Properties in Isometric Elbow Flexion
This study investigates age-related differences in motor unit (MU) properties and neuromuscular control during isometric elbow flexion across the human lifespan. High-density surface electromyography (sEMG) was recorded from the biceps brachii of 44 participants, divided into three groups: Child (8–14 years), Adult (20–40 years), and Elder (65–80 years). MU spike trains were extracted noninvasively by sEMG decomposition. Then the discharge rate, MU action potential (MUAP) morphology, recruitment threshold, and common neural drive were quantified and compared across age groups. This study provides novel insights into force tracking performance, revealing that both children and elders exhibit higher errors compared to young adults, likely due to immature or declining motor control systems. Significant differences in MU discharge patterns were observed across force levels and age groups. Children and elders displayed lower MU discharge rates at low force levels, which increased at higher forces. In contrast, adults demonstrated higher MU action potential peak-to-peak amplitudes (PPV) and recruitment thresholds (RTs), along with steeper PPV-RT slopes, suggesting a narrower RT range in children and older adults. Principal component analysis revealed a strong correlation between common neural drive and force across all groups, with neural drive being weaker in elders. Overall, young adults exhibited the most efficient and synchronized MU control, while children and older adults showed distinct deviations in discharge intensity, recruitment strategies, and neural synergy. These findings comprehensively characterize MU adaptations across the lifespan, offering implications for developmental neurophysiology and age-specific neuromuscular diagnostics and interventions.
- 12 August 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (8) , 869
Open AccessEffects of Aging on Motor Unit Properties in Isometric Elbow Flexion
This study investigates age-related differences in motor unit (MU) properties and neuromuscular control during isometric elbow flexion across the human lifespan. High-density surface electromyography (sEMG) was recorded from the biceps brachii of 44 participants, divided into three groups: Child (8–14 years), Adult (20–40 years), and Elder (65–80 years). MU spike trains were extracted noninvasively by sEMG decomposition. Then the discharge rate, MU action potential (MUAP) morphology, recruitment threshold, and common neural drive were quantified and compared across age groups. This study provides novel insights into force tracking performance, revealing that both children and elders exhibit higher errors compared to young adults, likely due to immature or declining motor control systems. Significant differences in MU discharge patterns were observed across force levels and age groups. Children and elders displayed lower MU discharge rates at low force levels, which increased at higher forces. In contrast, adults demonstrated higher MU action potential peak-to-peak amplitudes (PPV) and recruitment thresholds (RTs), along with steeper PPV-RT slopes, suggesting a narrower RT range in children and older adults. Principal component analysis revealed a strong correlation between common neural drive and force across all groups, with neural drive being weaker in elders. Overall, young adults exhibited the most efficient and synchronized MU control, while children and older adults showed distinct deviations in discharge intensity, recruitment strategies, and neural synergy. These findings comprehensively characterize MU adaptations across the lifespan, offering implications for developmental neurophysiology and age-specific neuromuscular diagnostics and interventions.
- Open Access
- 7 October 2022
- Published by Frontiers Media SA in Frontiers in Physiology
- Vol. 13, 956639
Are the antagonist muscle fatigued during a prolonged isometric fatiguing elbow flexion at very low forces for young adults?
The aim of this study was to examine whether antagonist muscles may be fatigued during a prolonged isometric fatiguing elbow flexion at very low forces. Twelve healthy male subjects sustained an isometric elbow flexion at 10% maximal voluntary contraction torque until exhaustion while multichannel electromyographic signals were collected from the biceps brachii (BB) and triceps brachii (TB). Muscle fiber conduction velocity (CV) and fractal dimension (FD) of both muscles were calculated to reflect peripheral and central fatigue. CV and FD of TB as well as FD of BB decreased progressively during the sustained fatiguing contraction, while the CV of BB declined at the beginning of the contraction and then increased progressively until the end of the contraction. The result may indicate that during the sustained low-force isometric fatiguing contraction, antagonist muscle may be peripherally fatigued, and changes in coactivation activities were modulated not only by central neuronal mechanisms of common drive but also by peripheral metabolic factors.
- 7 October 2022
- Published by Frontiers Media SA in Frontiers in Physiology
- Vol. 13, 956639
Open AccessAre the antagonist muscle fatigued during a prolonged isometric fatiguing elbow flexion at very low forces for young adults?
The aim of this study was to examine whether antagonist muscles may be fatigued during a prolonged isometric fatiguing elbow flexion at very low forces. Twelve healthy male subjects sustained an isometric elbow flexion at 10% maximal voluntary contraction torque until exhaustion while multichannel electromyographic signals were collected from the biceps brachii (BB) and triceps brachii (TB). Muscle fiber conduction velocity (CV) and fractal dimension (FD) of both muscles were calculated to reflect peripheral and central fatigue. CV and FD of TB as well as FD of BB decreased progressively during the sustained fatiguing contraction, while the CV of BB declined at the beginning of the contraction and then increased progressively until the end of the contraction. The result may indicate that during the sustained low-force isometric fatiguing contraction, antagonist muscle may be peripherally fatigued, and changes in coactivation activities were modulated not only by central neuronal mechanisms of common drive but also by peripheral metabolic factors.