Zeyu Liu
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
Research Interests
Zeyu Liu has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Long Cheng
School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China
6 shared publications - Zhengwei Li
School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China
4 shared publications - Houcheng Li
University of Chinese Academy of Sciences
3 shared publications - Haoyu Zhang
School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China
2 shared publications Jiachen Wei
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing, China
2 shared publications- Edward Wright
Faculty of Engineering, University of Nottingham, Nottingham NG7 2RD, UK
1 shared publication
Latest Publications
- 11 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 6350-6362
Design and Systematic Assessment of a Novel Underactuated Assistive Hand Exoskeleton
This study presents an innovative hand exoskeleton based on a linkage-transmission mechanism, aimed at supporting individuals with motor impairments in executing grasping activities essential for daily life. An underactuated mechanism leveraging redundant degrees of freedom is proposed, enabling the exoskeleton to adaptively grasp objects of varying shapes and sizes. The mechanical design ensures kinematic compatibility between the finger joints and the exoskeleton by establishing appropriate constrained kinematic chains. Its kinematics and quasi-statics are systematically analyzed to optimize the mechanical design. To evaluate the functional performance of the assistive hand exoskeleton, a systematic assessment framework is developed, incorporating three dimensions: subjective perception, kinesiology, and physiological response. Two experimental paradigms were conducted using the hand exoskeleton prototype with six healthy participants. Six assessment metrics (motion transparency, grip similarity, muscle activation, muscle fatigue, the System Usability Scale (SUS), and the Rating of Perceived Exertion (RPE) scale) were employed to assess the exoskeleton’s functionality. Experimental results demonstrate that the proposed exoskeleton effectively aids users in grasping various objects while significantly reducing physical exertion and fatigue.
- 11 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 6350-6362
Design and Systematic Assessment of a Novel Underactuated Assistive Hand Exoskeleton
This study presents an innovative hand exoskeleton based on a linkage-transmission mechanism, aimed at supporting individuals with motor impairments in executing grasping activities essential for daily life. An underactuated mechanism leveraging redundant degrees of freedom is proposed, enabling the exoskeleton to adaptively grasp objects of varying shapes and sizes. The mechanical design ensures kinematic compatibility between the finger joints and the exoskeleton by establishing appropriate constrained kinematic chains. Its kinematics and quasi-statics are systematically analyzed to optimize the mechanical design. To evaluate the functional performance of the assistive hand exoskeleton, a systematic assessment framework is developed, incorporating three dimensions: subjective perception, kinesiology, and physiological response. Two experimental paradigms were conducted using the hand exoskeleton prototype with six healthy participants. Six assessment metrics (motion transparency, grip similarity, muscle activation, muscle fatigue, the System Usability Scale (SUS), and the Rating of Perceived Exertion (RPE) scale) were employed to assess the exoskeleton’s functionality. Experimental results demonstrate that the proposed exoskeleton effectively aids users in grasping various objects while significantly reducing physical exertion and fatigue.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75, 9538512
A Fingertip Pose Reconstruction Method Based on Fiber Optic Bending Sensing
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75, 9538512
A Fingertip Pose Reconstruction Method Based on Fiber Optic Bending Sensing
- 30 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 30 (6) , 7809-7819
A Compact Dual-Core Waveguide Bending Sensor for Embodied Joint Angle Measurement
Bending sensors are extensively employed for joint angle measurement across various applications, including human–machine interaction (HMI), embodied intelligent systems, sports performance analysis, and rehabilitation therapy. Compared to video-based methods, wearable soft bending sensors offer broader applicability and are immune to occlusion. This article presents a compact dual-core waveguide bending sensor specifically designed for embodied joint angle measurement. Building upon prior tri-core research, the enhanced sensor leverages optical frequency for multichannel signal coupling and decoupling, enabling a more compact design. This advancement facilitates finger joint angle detection while significantly improving measurement capabilities. The proposed sensor consists of two detection core segments with distinct optical characteristics. Each segment contains two semi-cores of different colors to determine the bending direction. This design achieves compactness and reduces power consumption. The sensor exhibits excellent linearity, remarkable stability, and an average sensitivity of 0.17 dB/deg over a range of 0$^\circ$ to 90$^\circ$ in both directions. In addition, a bidirectional mamba neural network (BMNN) light signal decoupling algorithm is developed to map short-time optical signals to joint angles, achieving a cross-subject joint angle measurement error of 0.4$^\circ$. This advancement significantly enhances existing joint angle sensors, addressing challenges of reliability and accuracy in joint angle measurement. It holds the potential to advance fields reliant on precise human movement analysis, facilitating more accurate HMIs and the development of universally applicable embodied intelligent systems.
- 30 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 30 (6) , 7809-7819
A Compact Dual-Core Waveguide Bending Sensor for Embodied Joint Angle Measurement
Bending sensors are extensively employed for joint angle measurement across various applications, including human–machine interaction (HMI), embodied intelligent systems, sports performance analysis, and rehabilitation therapy. Compared to video-based methods, wearable soft bending sensors offer broader applicability and are immune to occlusion. This article presents a compact dual-core waveguide bending sensor specifically designed for embodied joint angle measurement. Building upon prior tri-core research, the enhanced sensor leverages optical frequency for multichannel signal coupling and decoupling, enabling a more compact design. This advancement facilitates finger joint angle detection while significantly improving measurement capabilities. The proposed sensor consists of two detection core segments with distinct optical characteristics. Each segment contains two semi-cores of different colors to determine the bending direction. This design achieves compactness and reduces power consumption. The sensor exhibits excellent linearity, remarkable stability, and an average sensitivity of 0.17 dB/deg over a range of 0$^\circ$ to 90$^\circ$ in both directions. In addition, a bidirectional mamba neural network (BMNN) light signal decoupling algorithm is developed to map short-time optical signals to joint angles, achieving a cross-subject joint angle measurement error of 0.4$^\circ$. This advancement significantly enhances existing joint angle sensors, addressing challenges of reliability and accuracy in joint angle measurement. It holds the potential to advance fields reliant on precise human movement analysis, facilitating more accurate HMIs and the development of universally applicable embodied intelligent systems.
- 12 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (8) , 2385-2396
A Fuzzy-Encoded Dual-Modal Soft Glove for Gesture and Grasping Object Classification
Human-machine interaction technologies are crucial for enhancing human capabilities in the digital world. The hand, as a primary interaction tool, conveys information through gestures and tactile signals. With advancements in flexible electronics and fuzzy systems, it is now possible to achieve interpretable and accurate interactions using hand gestures and tactile information. This paper introduces a fuzzy-encoded, dual-modal soft glove applied to gesture and grasping object classification. The glove, featuring 10 soft pressure sensors and 6 soft bending sensors, is compactly designed and exhibits excellent sensitivity, with average sensitivities of 133 N$^{-1}$ for soft pressure sensors and 13deg$^{-1}$ for soft bending sensors. The Fuzzy Mamba Encoding (FuME) classification algorithm, inspired by the interpretability and uncertainty resilience of fuzzy systems and the fitting ability of state-space models, was developed. To the best of our knowledge, this work is the first to combine the Mamba structure with fuzzy intelligent systems. Applied to gesture and grasping object classifications, the glove achieved average accuracies of 96.3% and 94.9%, respectively. The fuzzy encoding leverages membership vectors provided by experts to enhance interpretability. These results demonstrate the glove's effectiveness in signal recording and processing and highlight the powerful synergy between flexible electronic technologies and fuzzy systems.
- 12 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (8) , 2385-2396
A Fuzzy-Encoded Dual-Modal Soft Glove for Gesture and Grasping Object Classification
Human-machine interaction technologies are crucial for enhancing human capabilities in the digital world. The hand, as a primary interaction tool, conveys information through gestures and tactile signals. With advancements in flexible electronics and fuzzy systems, it is now possible to achieve interpretable and accurate interactions using hand gestures and tactile information. This paper introduces a fuzzy-encoded, dual-modal soft glove applied to gesture and grasping object classification. The glove, featuring 10 soft pressure sensors and 6 soft bending sensors, is compactly designed and exhibits excellent sensitivity, with average sensitivities of 133 N$^{-1}$ for soft pressure sensors and 13deg$^{-1}$ for soft bending sensors. The Fuzzy Mamba Encoding (FuME) classification algorithm, inspired by the interpretability and uncertainty resilience of fuzzy systems and the fitting ability of state-space models, was developed. To the best of our knowledge, this work is the first to combine the Mamba structure with fuzzy intelligent systems. Applied to gesture and grasping object classifications, the glove achieved average accuracies of 96.3% and 94.9%, respectively. The fuzzy encoding leverages membership vectors provided by experts to enhance interpretability. These results demonstrate the glove's effectiveness in signal recording and processing and highlight the powerful synergy between flexible electronic technologies and fuzzy systems.
- 1 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 23 (18) , 21821-21831
A Pressure-Based High-Quality Wrist Pulse Measurement System With an Application in Emotion Analysis
Wrist pulse contains much information about human beings, a typical pulse wave contains four significant morphological features which are related to the cardiovascular system and can be used for disease diagnosis. A high-quality wrist pulse measurement system is essential for the medical application of wrist pulses. In this work, a high-quality wrist pulse measurement system based on a pressure sensor is proposed. The measurement system has a maximum signal-to-noise ratio (SNR) of 53.44 dB and a satisfactory ability to resist environmental interferences. The system is applied to measure the wrist pulse signals of seven participants under different emotions and the signal quality is assessed in three signal quality indices (SQIs): the skewness, the kurtosis, and the perfusion index. The collected dataset presents better performance in all three SQIs than the DEAP dataset. The measured high-quality wrist pulse data are applied in emotion classification by an LSTM-GRU Boosting neural network. The emotion classification accuracy is 89.71% in valence and 80.06% in arousal. We also find features extracted by the proposed neural network have a strong connection with three morphological features of the wrist pulse, and we show the medical implication of these three morphological features with emotion.
- 1 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 23 (18) , 21821-21831
A Pressure-Based High-Quality Wrist Pulse Measurement System With an Application in Emotion Analysis
Wrist pulse contains much information about human beings, a typical pulse wave contains four significant morphological features which are related to the cardiovascular system and can be used for disease diagnosis. A high-quality wrist pulse measurement system is essential for the medical application of wrist pulses. In this work, a high-quality wrist pulse measurement system based on a pressure sensor is proposed. The measurement system has a maximum signal-to-noise ratio (SNR) of 53.44 dB and a satisfactory ability to resist environmental interferences. The system is applied to measure the wrist pulse signals of seven participants under different emotions and the signal quality is assessed in three signal quality indices (SQIs): the skewness, the kurtosis, and the perfusion index. The collected dataset presents better performance in all three SQIs than the DEAP dataset. The measured high-quality wrist pulse data are applied in emotion classification by an LSTM-GRU Boosting neural network. The emotion classification accuracy is 89.71% in valence and 80.06% in arousal. We also find features extracted by the proposed neural network have a strong connection with three morphological features of the wrist pulse, and we show the medical implication of these three morphological features with emotion.
- 16 November 2021
- Published by Wiley in Asia-Pacific Journal of Chemical Engineering
- Vol. 17 (1) , e2719
Development of a novel artificial neural network model for closed pulsating heat pipe with water and aqueous solutions
Aqueous solutions are increasingly introduced as the working fluids of pulsating heat pipe (PHP), to balance and optimise certain thermo‐physical properties of pure water. In this paper, a first attempt is conducted to predict the heat transfer performance of closed PHP with pure water and available aqueous solutions based on a fully connected feed‐forward artificial neural network (ANN). The notable dimensionless numbers, Kutateladze number (Ku), Bond number (Bo), Morton number (Mo), Jackob number (Ja), Prandtl number (Pr), Laplace number (La), d/Le and number of turns (N) were selected as the inputs of the ANN model. It was found that the prediction of proposed ANN model posed a good agreement with experimental data with the MSE and correlation coefficient of 0.026 and 0.979, respectively. For 81.02% of the collected data, the absolute deviation of prediction for thermal resistance was within 25%. For 95% of collected data, the prediction of thermal resistance and temperature difference between the evaporation and condensation section fell within ±0.37 K/W, and ±17.1 K, respectively. The evaluation of evaporation and condensation section temperatures was also presented based on the proposed ANN model.
- 16 November 2021
- Published by Wiley in Asia-Pacific Journal of Chemical Engineering
- Vol. 17 (1) , e2719
Development of a novel artificial neural network model for closed pulsating heat pipe with water and aqueous solutions
Aqueous solutions are increasingly introduced as the working fluids of pulsating heat pipe (PHP), to balance and optimise certain thermo‐physical properties of pure water. In this paper, a first attempt is conducted to predict the heat transfer performance of closed PHP with pure water and available aqueous solutions based on a fully connected feed‐forward artificial neural network (ANN). The notable dimensionless numbers, Kutateladze number (Ku), Bond number (Bo), Morton number (Mo), Jackob number (Ja), Prandtl number (Pr), Laplace number (La), d/Le and number of turns (N) were selected as the inputs of the ANN model. It was found that the prediction of proposed ANN model posed a good agreement with experimental data with the MSE and correlation coefficient of 0.026 and 0.979, respectively. For 81.02% of the collected data, the absolute deviation of prediction for thermal resistance was within 25%. For 95% of collected data, the prediction of thermal resistance and temperature difference between the evaporation and condensation section fell within ±0.37 K/W, and ±17.1 K, respectively. The evaluation of evaporation and condensation section temperatures was also presented based on the proposed ANN model.