Lizheng Pan
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
Lizheng Pan has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- Aiguo Song
Southeast University
6 shared publications Suolin Duan
School of Mechanical Engineering, Changzhou University, Changzhou, People’s Republic of China
6 shared publicationsAiguo Song
School of Instrument Science and Engineering, Southeast University, Nanjing, China
4 shared publicationsDashuai Zhu
School of Mechanical Engineering, Changzhou University, Changzhou, P.R. China
4 shared publicationsShigang She
School of Mechanical Engineering, Changzhou University, Changzhou, P.R. China
4 shared publicationsXianchuan Shi
School of Mechanical Engineering, Changzhou University, Changzhou, P.R. China
4 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (17) , 8775
Weft Skew Detection in Woven Fabrics Based on Rotation-Based Projection Guidance and Adaptive Frequency-Domain Reconstruction
Conventional weft skew detection methods are often sensitive to edge quality and parameter settings and may perform inconsistently across fabrics with different weave structures, densities, local texture disturbances, and interlaced warp–weft textures. This study proposes a coarse-to-fine method combining rotation-based projection guidance with adaptive frequency-domain reconstruction. Rotation-based projection profiling first provides a coarse-angle constraint. The dominant-frequency radius and dominant spectral orientation are then determined adaptively to construct a frequency-domain mask for weft-yarn texture reconstruction. Candidate weft-yarn line segments are extracted by local adaptive thresholding, skeleton extraction, line fitting, and screening. The final angle is obtained using a length-weighted circular mean and median absolute deviation (MAD)-based outlier rejection. The method was evaluated on 164 distinct fabric samples comprising 1804 test images. It achieved a mean absolute error (MAE) of 0.09°, a root mean square error (RMSE) of 0.12°, a maximum absolute error of 0.48°, and an average runtime of 104.36 ms. Compared with rotation-based projection profiling, the Hough transform, and a baseline method combining the fast Fourier transform (FFT) with the Hough transform, the proposed method yielded lower errors and a more concentrated error distribution, supporting its effectiveness for automated weft skew detection in woven fabrics.
- 3 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (17) , 8775
Open AccessWeft Skew Detection in Woven Fabrics Based on Rotation-Based Projection Guidance and Adaptive Frequency-Domain Reconstruction
Conventional weft skew detection methods are often sensitive to edge quality and parameter settings and may perform inconsistently across fabrics with different weave structures, densities, local texture disturbances, and interlaced warp–weft textures. This study proposes a coarse-to-fine method combining rotation-based projection guidance with adaptive frequency-domain reconstruction. Rotation-based projection profiling first provides a coarse-angle constraint. The dominant-frequency radius and dominant spectral orientation are then determined adaptively to construct a frequency-domain mask for weft-yarn texture reconstruction. Candidate weft-yarn line segments are extracted by local adaptive thresholding, skeleton extraction, line fitting, and screening. The final angle is obtained using a length-weighted circular mean and median absolute deviation (MAD)-based outlier rejection. The method was evaluated on 164 distinct fabric samples comprising 1804 test images. It achieved a mean absolute error (MAE) of 0.09°, a root mean square error (RMSE) of 0.12°, a maximum absolute error of 0.48°, and an average runtime of 104.36 ms. Compared with rotation-based projection profiling, the Hough transform, and a baseline method combining the fast Fourier transform (FFT) with the Hough transform, the proposed method yielded lower errors and a more concentrated error distribution, supporting its effectiveness for automated weft skew detection in woven fabrics.
- 3 December 2025
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (1) , 015005
BrainEmoNet: emotion recognition network based on brain function asymmetry
The recognition of the subject’s emotional states is of great significance for achieving humanized services in many scenarios with human-computer interaction. Recently, identification of the emotional states based on electroencephalogram (EEG) has received increasing attention. However, due to the complexity of EEG signals, EEG-based emotion recognition is very challenging. In this research, a novel BrainEmoNet with learning-based framework is proposed to improve the emotion recognition accuracy from the perspective of the asymmetry of human brain functions. The BrainEmoNet consists of frequency-domain feature network (FFN), long-term dependent feature network (LDFN) and spatial characteristic analysis network (SCAN). The parallel FFN and LDFN are suggested to extract the frequency-domain and long-term dependent features of the information in each brain channel, respectively. Meanwhile, based on the working principle of the human brain, the SCAN with channel-spatial attention mechanism is proposed to focus on the high-value information channels with assigning adaptive weights and analyze the spatial characteristics of the frequency-domain and time-domain features. The feature analysis in the time-frequency-spatial perspective can fully explore the emotional information contained in EEG information. Experimental results on multi-modal DEAP dataset presents the competitive performances of the BrainEmoNet over the existing state-of-the-art models. In the subject-dependent experiments, the proposed model achieves identification accuracies of 86.77% and 82.14% in arousal and valence dimensions, respectively, compared to 75.53% and 72.83% in the subject-independent experiments. The proposed BrainEmoNet model in this research can be used as an auxiliary tool for the assessment or monitoring of emotions.
- 3 December 2025
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (1) , 015005
BrainEmoNet: emotion recognition network based on brain function asymmetry
The recognition of the subject’s emotional states is of great significance for achieving humanized services in many scenarios with human-computer interaction. Recently, identification of the emotional states based on electroencephalogram (EEG) has received increasing attention. However, due to the complexity of EEG signals, EEG-based emotion recognition is very challenging. In this research, a novel BrainEmoNet with learning-based framework is proposed to improve the emotion recognition accuracy from the perspective of the asymmetry of human brain functions. The BrainEmoNet consists of frequency-domain feature network (FFN), long-term dependent feature network (LDFN) and spatial characteristic analysis network (SCAN). The parallel FFN and LDFN are suggested to extract the frequency-domain and long-term dependent features of the information in each brain channel, respectively. Meanwhile, based on the working principle of the human brain, the SCAN with channel-spatial attention mechanism is proposed to focus on the high-value information channels with assigning adaptive weights and analyze the spatial characteristics of the frequency-domain and time-domain features. The feature analysis in the time-frequency-spatial perspective can fully explore the emotional information contained in EEG information. Experimental results on multi-modal DEAP dataset presents the competitive performances of the BrainEmoNet over the existing state-of-the-art models. In the subject-dependent experiments, the proposed model achieves identification accuracies of 86.77% and 82.14% in arousal and valence dimensions, respectively, compared to 75.53% and 72.83% in the subject-independent experiments. The proposed BrainEmoNet model in this research can be used as an auxiliary tool for the assessment or monitoring of emotions.
- 20 November 2025
- Published by IOP Publishing in Engineering Research Express
- Vol. 7 (4) , 045286
Multi-scale feature fusion network-based weft density analysis
The detection of weft density in fabrics is an important indicator for fabric quality control. Traditional methods for weft density analysis based on machine vision require the design according to fabric characteristics, lacking adaptability for different fabrics. To address this issue, a weft density detection method based on a multi-scale feature fusion network is proposed. First, a fabric image acquisition system with a transmissive light source is constructed to create a dataset with weft yarn contours. Second, the multi-scale feature fusion network (MSRUNeXt) is designed, which achieves precise segmentation and feature extraction of fabric weft textures through the collaborative cooperation of the multi-scale residual convolution module (MRCM), dual-channel upsampling module (DCUM), and global-local spatial attention module (GLSA). Finally, the progressive probabilistic Hough transform method is used to calculate the weft inclination angle and correct the fabric image, and the spatial positions of weft yarns are statistically analyzed by gray projection methods to realize precise calculation of weft density. Experimental results demonstrate that this method effectively detects weft density in different fabric types, exhibiting good adaptability and accuracy.
- 20 November 2025
- Published by IOP Publishing in Engineering Research Express
- Vol. 7 (4) , 045286
Multi-scale feature fusion network-based weft density analysis
The detection of weft density in fabrics is an important indicator for fabric quality control. Traditional methods for weft density analysis based on machine vision require the design according to fabric characteristics, lacking adaptability for different fabrics. To address this issue, a weft density detection method based on a multi-scale feature fusion network is proposed. First, a fabric image acquisition system with a transmissive light source is constructed to create a dataset with weft yarn contours. Second, the multi-scale feature fusion network (MSRUNeXt) is designed, which achieves precise segmentation and feature extraction of fabric weft textures through the collaborative cooperation of the multi-scale residual convolution module (MRCM), dual-channel upsampling module (DCUM), and global-local spatial attention module (GLSA). Finally, the progressive probabilistic Hough transform method is used to calculate the weft inclination angle and correct the fabric image, and the spatial positions of weft yarns are statistically analyzed by gray projection methods to realize precise calculation of weft density. Experimental results demonstrate that this method effectively detects weft density in different fabric types, exhibiting good adaptability and accuracy.
- Open Access
- 28 January 2025
- Published by MDPI AG in Brain Sciences
- Vol. 15 (2) , 129
MSEI-ENet: A Multi-Scale EEG-Inception Integrated Encoder Network for Motor Imagery EEG Decoding
Background: Due to complex signal characteristics and distinct individual differences, the decoding of a motor imagery electroencephalogram (MI-EEG) is limited by the unsatisfactory performance of suboptimal traditional models. Methods: A subject-independent model named MSEI-ENet is proposed for multiple-task MI-EEG decoding. It employs a specially designed multi-scale structure EEG-inception module (MSEI) for comprehensive feature learning. The encoder module further helps to detect discriminative information by its multi-head self-attention layer with a larger receptive field, which enhances feature representation and improves recognition efficacy. Results: The experimental results on Competition IV dataset 2a showed that our proposed model yielded an overall accuracy of 94.30%, MF1 score of 94.31%, and Kappa of 0.92. Conclusions: A performance comparison with state-of-the-art methods demonstrated the effectiveness and generalizability of the proposed model on challenging multi-task MI-EEG decoding.
- 28 January 2025
- Published by MDPI AG in Brain Sciences
- Vol. 15 (2) , 129
Open AccessMSEI-ENet: A Multi-Scale EEG-Inception Integrated Encoder Network for Motor Imagery EEG Decoding
Background: Due to complex signal characteristics and distinct individual differences, the decoding of a motor imagery electroencephalogram (MI-EEG) is limited by the unsatisfactory performance of suboptimal traditional models. Methods: A subject-independent model named MSEI-ENet is proposed for multiple-task MI-EEG decoding. It employs a specially designed multi-scale structure EEG-inception module (MSEI) for comprehensive feature learning. The encoder module further helps to detect discriminative information by its multi-head self-attention layer with a larger receptive field, which enhances feature representation and improves recognition efficacy. Results: The experimental results on Competition IV dataset 2a showed that our proposed model yielded an overall accuracy of 94.30%, MF1 score of 94.31%, and Kappa of 0.92. Conclusions: A performance comparison with state-of-the-art methods demonstrated the effectiveness and generalizability of the proposed model on challenging multi-task MI-EEG decoding.
- 20 March 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 173, 108300
A sleep staging model on wavelet-based adaptive spectrogram reconstruction and light weight CNN
- 20 March 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 173, 108300
A sleep staging model on wavelet-based adaptive spectrogram reconstruction and light weight CNN
- 1 December 2023
- Published by IOP Publishing in Physiological Measurement
- Vol. 44 (12) , 125006
Cross-subject emotion recognition using hierarchical feature optimization and support vector machine with multi-kernel collaboration
Objective . Due to individual differences, it is greatly challenging to realize the multiple types of emotion identification across subjects. Approach . In this research, a hierarchical feature optimization method is proposed in order to represent emotional states effectively based on peripheral physiological signals. Firstly, sparse learning combined with binary search is employed to achieve feature selection of single signals. Then an improved fast correlation-based filter is proposed to implement fusion optimization of multi-channel signal features. Aiming at overcoming the limitations of the support vector machine (SVM), which uses a single kernel function to make decisions, the multi-kernel function collaboration strategy is proposed to improve the classification performance of SVM. Main results. The effectiveness of the proposed method is verified on the DEAP dataset. Experimental results show that the proposed method presents a competitive performance for four cross-subject types of emotion identification with an accuracy of 84% (group 1) and 85.07% (group 2). Significance . The proposed model with hierarchical feature optimization and SVM with multi-kernel function collaboration demonstrates superior emotion recognition accuracy compared to state-of-the-art techniques. In addition, the analysis based on DEAP dataset composition characteristics presents a novel perspective to explore the emotion recognition issue more objectively and comprehensively.
- 1 December 2023
- Published by IOP Publishing in Physiological Measurement
- Vol. 44 (12) , 125006
Cross-subject emotion recognition using hierarchical feature optimization and support vector machine with multi-kernel collaboration
Objective . Due to individual differences, it is greatly challenging to realize the multiple types of emotion identification across subjects. Approach . In this research, a hierarchical feature optimization method is proposed in order to represent emotional states effectively based on peripheral physiological signals. Firstly, sparse learning combined with binary search is employed to achieve feature selection of single signals. Then an improved fast correlation-based filter is proposed to implement fusion optimization of multi-channel signal features. Aiming at overcoming the limitations of the support vector machine (SVM), which uses a single kernel function to make decisions, the multi-kernel function collaboration strategy is proposed to improve the classification performance of SVM. Main results. The effectiveness of the proposed method is verified on the DEAP dataset. Experimental results show that the proposed method presents a competitive performance for four cross-subject types of emotion identification with an accuracy of 84% (group 1) and 85.07% (group 2). Significance . The proposed model with hierarchical feature optimization and SVM with multi-kernel function collaboration demonstrates superior emotion recognition accuracy compared to state-of-the-art techniques. In addition, the analysis based on DEAP dataset composition characteristics presents a novel perspective to explore the emotion recognition issue more objectively and comprehensively.
- 24 June 2022
- Published by Springer Nature in International Journal of Machine Learning and Cybernetics
- Vol. 13 (10) , 3225-3237
A universal emotion recognition method based on feature priority evaluation and classifier reinforcement
- 24 June 2022
- Published by Springer Nature in International Journal of Machine Learning and Cybernetics
- Vol. 13 (10) , 3225-3237
A universal emotion recognition method based on feature priority evaluation and classifier reinforcement
- 27 November 2020
- Published by Taylor & Francis in Australian Journal of Mechanical Engineering
- Vol. 21 (2) , 428-441
Research on feature fusion strategy for gear states diagnosis based on fusion assessment
Gear is commonly applied as a vital component that connects and transmits power, which plays a crucial role in mechanical system. In order to represent the running states of the gear effectively and improve the diagnosis accuracy further, a feature fusion strategy based on fusion assessment to obtain optimum feature fusion pattern is proposed in this research, including two-direction feature fusion strategy and three-direction feature fusion strategy. Firstly, different feature extraction methods are employed to extract the features of vibration signal in each direction (X, Y and Z axis), respectively. Then, the best fusion mode is determined by fusion assessment mechanism based on fuzzy logic in two directions and vibration intensity of signal in three directions. Finally, support vector machine (SVM) and decision tree (DT) are selected to verify the validity and universality of the proposed method. Experimental studies show that feature fusion strategy for gear states diagnosis based on fusion assessment can take full advantage of the complementary performances of different feature extraction methods and signal characteristics in different directions, which can fully represent the health states of the running gear and effectively improve the diagnosis accuracy.
- 27 November 2020
- Published by Taylor & Francis in Australian Journal of Mechanical Engineering
- Vol. 21 (2) , 428-441
Research on feature fusion strategy for gear states diagnosis based on fusion assessment
Gear is commonly applied as a vital component that connects and transmits power, which plays a crucial role in mechanical system. In order to represent the running states of the gear effectively and improve the diagnosis accuracy further, a feature fusion strategy based on fusion assessment to obtain optimum feature fusion pattern is proposed in this research, including two-direction feature fusion strategy and three-direction feature fusion strategy. Firstly, different feature extraction methods are employed to extract the features of vibration signal in each direction (X, Y and Z axis), respectively. Then, the best fusion mode is determined by fusion assessment mechanism based on fuzzy logic in two directions and vibration intensity of signal in three directions. Finally, support vector machine (SVM) and decision tree (DT) are selected to verify the validity and universality of the proposed method. Experimental studies show that feature fusion strategy for gear states diagnosis based on fusion assessment can take full advantage of the complementary performances of different feature extraction methods and signal characteristics in different directions, which can fully represent the health states of the running gear and effectively improve the diagnosis accuracy.
- Open Access
Emotional State Recognition from Peripheral Physiological Signals Using Fused Nonlinear Features and Team-Collaboration Identification Strategy
Emotion recognition realizing human inner perception has a very important application prospect in human-computer interaction. In order to improve the accuracy of emotion recognition, a novel method combining fused nonlinear features and team-collaboration identification strategy was proposed for emotion recognition using physiological signals. Four nonlinear features, namely approximate entropy (ApEn), sample entropy (SaEn), fuzzy entropy (FuEn) and wavelet packet entropy (WpEn) are employed to reflect emotional states deeply with each type of physiological signal. Then the features of different physiological signals are fused to represent the emotional states from multiple perspectives. Each classifier has its own advantages and disadvantages. In order to make full use of the advantages of other classifiers and avoid the limitation of single classifier, the team-collaboration model is built and the team-collaboration decision-making mechanism is designed according to the proposed team-collaboration identification strategy which is based on the fusion of support vector machine (SVM), decision tree (DT) and extreme learning machine (ELM). Through analysis, SVM is selected as the main classifier with DT and ELM as auxiliary classifiers. According to the designed decision-making mechanism, the proposed team-collaboration identification strategy can effectively employ different classification methods to make decision based on the characteristics of the samples through SVM classification. For samples which are easy to be identified by SVM, SVM directly determines the identification results, whereas SVM-DT-ELM collaboratively determines the identification results, which can effectively utilize the characteristics of each classifier and improve the classification accuracy. The effectiveness and universality of the proposed method are verified by Augsburg database and database for emotion analysis using physiological (DEAP) signals. The experimental results uniformly indicated that the proposed method combining fused nonlinear features and team-collaboration identification strategy presents better performance than the existing methods.
Open AccessEmotional State Recognition from Peripheral Physiological Signals Using Fused Nonlinear Features and Team-Collaboration Identification Strategy
Emotion recognition realizing human inner perception has a very important application prospect in human-computer interaction. In order to improve the accuracy of emotion recognition, a novel method combining fused nonlinear features and team-collaboration identification strategy was proposed for emotion recognition using physiological signals. Four nonlinear features, namely approximate entropy (ApEn), sample entropy (SaEn), fuzzy entropy (FuEn) and wavelet packet entropy (WpEn) are employed to reflect emotional states deeply with each type of physiological signal. Then the features of different physiological signals are fused to represent the emotional states from multiple perspectives. Each classifier has its own advantages and disadvantages. In order to make full use of the advantages of other classifiers and avoid the limitation of single classifier, the team-collaboration model is built and the team-collaboration decision-making mechanism is designed according to the proposed team-collaboration identification strategy which is based on the fusion of support vector machine (SVM), decision tree (DT) and extreme learning machine (ELM). Through analysis, SVM is selected as the main classifier with DT and ELM as auxiliary classifiers. According to the designed decision-making mechanism, the proposed team-collaboration identification strategy can effectively employ different classification methods to make decision based on the characteristics of the samples through SVM classification. For samples which are easy to be identified by SVM, SVM directly determines the identification results, whereas SVM-DT-ELM collaboratively determines the identification results, which can effectively utilize the characteristics of each classifier and improve the classification accuracy. The effectiveness and universality of the proposed method are verified by Augsburg database and database for emotion analysis using physiological (DEAP) signals. The experimental results uniformly indicated that the proposed method combining fused nonlinear features and team-collaboration identification strategy presents better performance than the existing methods.
- Open Access
- 1 September 2019
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 16 (5)
Study on motion performance of robot-aided passive rehabilitation exercises using novel dynamic motion planning strategy
The motion rehabilitation training robot is developed to help patients with motion dysfunction recover their motor function by providing a large amount of repetitive robot-aided exercise. To achieve stable and smooth robot-aided exercises for stroke patients, a motion control method with a novel dynamic motion planning strategy is proposed. The physical state of the training limb is assessed real time during the rehabilitation exercises. The dynamic motion planning strategy is developed by employing a suitable interpolation method dynamically corresponding to the physical state of the training limb to plan a trajectory tracking system that completely utilizes different interpolation characteristics to manage the movement in accordance with the time-varying physical state of the training limb. Concurrently, a position-based impedance control is adopted to achieve compliant movement. Functional (quantitative and qualitative) and clinical experiments are conducted on a four-degree-of-freedom whole-arm manipulator upper limb rehabilitation robot to verify the effectiveness of the control method designed with the dynamic motion planning strategy. The results indicate that the proposed control strategy can exhibit better performances in terms of the stability and smoothness.
- 1 September 2019
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 16 (5)
Open AccessStudy on motion performance of robot-aided passive rehabilitation exercises using novel dynamic motion planning strategy
The motion rehabilitation training robot is developed to help patients with motion dysfunction recover their motor function by providing a large amount of repetitive robot-aided exercise. To achieve stable and smooth robot-aided exercises for stroke patients, a motion control method with a novel dynamic motion planning strategy is proposed. The physical state of the training limb is assessed real time during the rehabilitation exercises. The dynamic motion planning strategy is developed by employing a suitable interpolation method dynamically corresponding to the physical state of the training limb to plan a trajectory tracking system that completely utilizes different interpolation characteristics to manage the movement in accordance with the time-varying physical state of the training limb. Concurrently, a position-based impedance control is adopted to achieve compliant movement. Functional (quantitative and qualitative) and clinical experiments are conducted on a four-degree-of-freedom whole-arm manipulator upper limb rehabilitation robot to verify the effectiveness of the control method designed with the dynamic motion planning strategy. The results indicate that the proposed control strategy can exhibit better performances in terms of the stability and smoothness.