Xiaowei Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiaowei Zhang has a total of 129 co-authors. Here are the top co-authors by number of shared publications:
- Bin Hu
Key Laboratory of Brain Health Intelligent Evaluation and Intervention, Ministry of Education, Beijing Institute of Technology, Beijing, China
23 shared publications - Jian Shen
School of Medical Technology, Beijing Institute of Technology, Beijing, China
12 shared publications - Kechen Hou
Gansu Provincial Key Laboratory of Wearable Computing, School of Information Science and Engineering, Lanzhou University, Lanzhou, China
7 shared publications - Qiqi Zhao
Lanzhou University, Lanzhou, China
5 shared publications - Dawei Lu
School of Optics and Electronics, Beijing Institute of Technology, Beijing 100081, China
4 shared publications - Rui Li
Gansu Provincial Key Laboratory of Wearable Computing, School of Information Science and Engineering, Lanzhou University, Lanzhou, China
4 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 16141-16150
Differential Mental Disorder Detection with Psychology-Inspired Multimodal Stimuli
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 16141-16150
Open AccessDifferential Mental Disorder Detection with Psychology-Inspired Multimodal Stimuli
- Open Access
- 24 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 4921
Knowledge-Refined Causality-Inspired Preference Modeling for Recommender Systems
Recommender systems have become a key technology for intelligent information services by learning user preferences from large-scale behavioral data and heterogeneous knowledge sources. However, existing recommendation methods still face two critical limitations: noisy knowledge graphs may introduce misleading signals during representation learning, and it remains challenging to identify item attributes that are relevant to users’ observed preferences. To address these issues, we propose CaRecK (causality-inspired recommendation with knowledge graph), a knowledge-refined preference modeling framework that combines causality-inspired attribute relevance estimation with preference-aware knowledge aggregation. Specifically, we develop an attribute selection mechanism to identify preference-relevant attributes for user representation learning. Furthermore, a preference-aware strategy is designed to adaptively select informative neighbors and integrate semantic information for learning item representations. Extensive experiments on two real-world benchmark datasets demonstrate that CaRecK achieves competitive performance in both CTR prediction and Top-K recommendation tasks.
- 24 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 4921
Open AccessKnowledge-Refined Causality-Inspired Preference Modeling for Recommender Systems
Recommender systems have become a key technology for intelligent information services by learning user preferences from large-scale behavioral data and heterogeneous knowledge sources. However, existing recommendation methods still face two critical limitations: noisy knowledge graphs may introduce misleading signals during representation learning, and it remains challenging to identify item attributes that are relevant to users’ observed preferences. To address these issues, we propose CaRecK (causality-inspired recommendation with knowledge graph), a knowledge-refined preference modeling framework that combines causality-inspired attribute relevance estimation with preference-aware knowledge aggregation. Specifically, we develop an attribute selection mechanism to identify preference-relevant attributes for user representation learning. Furthermore, a preference-aware strategy is designed to adaptively select informative neighbors and integrate semantic information for learning item representations. Extensive experiments on two real-world benchmark datasets demonstrate that CaRecK achieves competitive performance in both CTR prediction and Top-K recommendation tasks.
- Open Access
- 26 September 2026
- Published by MDPI AG in Electronics
- Vol. 15 (19) , 4438
Multimodal Decoupling Chain of Thought for Affective Understanding
The Chain-of-Thought (CoT) technique has significantly improved the affective reasoning capabilities of large language models (LLMs). However, existing multimodal CoT typically aggregate information from different modalities into mixed inputs, which allows the model to perform entangled reasoning. This paradigm often exhibits cross-modal information confusion and misalignment when dealing with complex sentiment samples with semantic conflicts. To address these challenges, we propose a Multimodal Decoupling Chain-of-Thought for affective understanding called MDCoT. Specifically, MDCoT structures LLM reasoning into three explicit and progressive steps: (1) Shared Sentiment Aggregation, which extracts common emotional baselines aligned across all modalities; (2) Private Sentiment Disentanglement, which isolates modality-specific emotional nuances, such as textual metaphors, facial micro-expressions, or audio tone shifts; and (3) Global Integrated Inference, which balances shared consensus and modal-private clues to make the final prediction. Extensive experiments across three affective tasks (i.e., sentiment analysis, sarcasm detection, intent recognition) demonstrate that MDCoT consistently outperforms state-of-the-art end-to-end models and existing prompting strategies across multiple MLLMs. Furthermore, ablation and robustness studies confirm that explicit disentanglement effectively prevents cross-modal conflict and improves decision stability.
- 26 September 2026
- Published by MDPI AG in Electronics
- Vol. 15 (19) , 4438
Open AccessMultimodal Decoupling Chain of Thought for Affective Understanding
The Chain-of-Thought (CoT) technique has significantly improved the affective reasoning capabilities of large language models (LLMs). However, existing multimodal CoT typically aggregate information from different modalities into mixed inputs, which allows the model to perform entangled reasoning. This paradigm often exhibits cross-modal information confusion and misalignment when dealing with complex sentiment samples with semantic conflicts. To address these challenges, we propose a Multimodal Decoupling Chain-of-Thought for affective understanding called MDCoT. Specifically, MDCoT structures LLM reasoning into three explicit and progressive steps: (1) Shared Sentiment Aggregation, which extracts common emotional baselines aligned across all modalities; (2) Private Sentiment Disentanglement, which isolates modality-specific emotional nuances, such as textual metaphors, facial micro-expressions, or audio tone shifts; and (3) Global Integrated Inference, which balances shared consensus and modal-private clues to make the final prediction. Extensive experiments across three affective tasks (i.e., sentiment analysis, sarcasm detection, intent recognition) demonstrate that MDCoT consistently outperforms state-of-the-art end-to-end models and existing prompting strategies across multiple MLLMs. Furthermore, ablation and robustness studies confirm that explicit disentanglement effectively prevents cross-modal conflict and improves decision stability.
- Top Cited Papers
- 12 February 2026
- Published by Elsevier in Information Fusion
- Vol. 132
Emotion recognition using multimodal physiological signals through regional to global fusion with a spatial-temporal semantic alignment mechanism
- 12 February 2026
- Published by Elsevier in Information Fusion
- Vol. 132
Top Cited PapersEmotion recognition using multimodal physiological signals through regional to global fusion with a spatial-temporal semantic alignment mechanism
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (2) , 1677-1694
SMA-EL:A Minimal 1-Cycle Construction Algorithm With Simplicial Maps Annotation and Edge Loss for Emotional Brain Networks Analysis
The brain patterns of emotional perception remain a pivotal research domain in affective neuroscience. Modeling the brain as a complex network has become a crucial approach to understanding its functions. However, traditional brain network research based on graph theory primarily focuses on dyadic interactions between brain regions, which cannot effectively characterize the information exchange process among multiple brain regions during emotional cognitive processes. To address these limitations, we shift our perspective from graph theory to the topological data analysis (TDA) of minimal 1-cycles. The 1 cycles or loops within a network represent the fundamental high order interactions in complex networks and serve as essential pathways for information transmission and integration among the distributed networks of brain regions. By focusing on cycle structures in affective brain networks, we propose a novel SMA-EL method based on the collaborative optimization of the Minimal 1-Cycle with Simplicial Maps Annotation (SMA-M1C) method and linear programming, which balances computational efficiency and method performance to reconstruct the optimal cycles in the brain network. This method is applied to the analysis of emotional brain networks in response to positive and negative emotions induced by naturalistic viewing. Comprehensive experiments demonstrate that the 1-cycle structures of the brain's functional patterns exhibit differences at both individual and group levels, aligning with prior research. Furthermore, the 1 cycles we proposed can serve as a biological marker for emotion recognition. These findings may provide new insights into the organization patterns of functional brain networks under diverse emotional states.
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (2) , 1677-1694
SMA-EL:A Minimal 1-Cycle Construction Algorithm With Simplicial Maps Annotation and Edge Loss for Emotional Brain Networks Analysis
The brain patterns of emotional perception remain a pivotal research domain in affective neuroscience. Modeling the brain as a complex network has become a crucial approach to understanding its functions. However, traditional brain network research based on graph theory primarily focuses on dyadic interactions between brain regions, which cannot effectively characterize the information exchange process among multiple brain regions during emotional cognitive processes. To address these limitations, we shift our perspective from graph theory to the topological data analysis (TDA) of minimal 1-cycles. The 1 cycles or loops within a network represent the fundamental high order interactions in complex networks and serve as essential pathways for information transmission and integration among the distributed networks of brain regions. By focusing on cycle structures in affective brain networks, we propose a novel SMA-EL method based on the collaborative optimization of the Minimal 1-Cycle with Simplicial Maps Annotation (SMA-M1C) method and linear programming, which balances computational efficiency and method performance to reconstruct the optimal cycles in the brain network. This method is applied to the analysis of emotional brain networks in response to positive and negative emotions induced by naturalistic viewing. Comprehensive experiments demonstrate that the 1-cycle structures of the brain's functional patterns exhibit differences at both individual and group levels, aligning with prior research. Furthermore, the 1 cycles we proposed can serve as a biological marker for emotion recognition. These findings may provide new insights into the organization patterns of functional brain networks under diverse emotional states.
- Top Cited Papers
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (1) , 1091-1104
WDANet: Wasserstein Distribution Inspired Dynamic Adversarial Network for EEG-Based Cross-Domain Depression Recognition
Researchers have long sought objective and quantifiable methods for recognizing depression. Electroencephalography (EEG) signals, which reflect brain activities objectively, have emerged as a promising tool for this purpose. However, the practical application of EEG signals faces significant challenges arising from distribution variability across different datasets and subjects. In addition, conventional methods often struggle to effectively capture information related to dynamic transformations in distributions. To address these issues, we propose a Wasserstein distribution-inspired dynamic adversarial network (WDANet) for EEG-based depression recognition. Specifically, WDANet includes a global discriminator that focuses on the marginal distribution of EEG features, a local discriminator that concentrates on the conditional distribution of EEG features, and a Wasserstein distribution discriminator that utilizes Wasserstein distributions derived from various processed EEG features. The experimental results show that WDANet achieved classification accuracies of 83.33%, 75.52%, 73.93%, 76.04%, and 70.94% in cross-subject, cross-dataset experiments conducted on three datasets, demonstrating its effectiveness and superiority compared to state-of-the-art methods. These results support our claim that WDANet enhances the accuracy and interpretability of depression recognition, providing insights and new research directions for the integration of neuroscience and artificial intelligence technologies.
- 19 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 17 (1) , 1091-1104
Top Cited PapersWDANet: Wasserstein Distribution Inspired Dynamic Adversarial Network for EEG-Based Cross-Domain Depression Recognition
Researchers have long sought objective and quantifiable methods for recognizing depression. Electroencephalography (EEG) signals, which reflect brain activities objectively, have emerged as a promising tool for this purpose. However, the practical application of EEG signals faces significant challenges arising from distribution variability across different datasets and subjects. In addition, conventional methods often struggle to effectively capture information related to dynamic transformations in distributions. To address these issues, we propose a Wasserstein distribution-inspired dynamic adversarial network (WDANet) for EEG-based depression recognition. Specifically, WDANet includes a global discriminator that focuses on the marginal distribution of EEG features, a local discriminator that concentrates on the conditional distribution of EEG features, and a Wasserstein distribution discriminator that utilizes Wasserstein distributions derived from various processed EEG features. The experimental results show that WDANet achieved classification accuracies of 83.33%, 75.52%, 73.93%, 76.04%, and 70.94% in cross-subject, cross-dataset experiments conducted on three datasets, demonstrating its effectiveness and superiority compared to state-of-the-art methods. These results support our claim that WDANet enhances the accuracy and interpretability of depression recognition, providing insights and new research directions for the integration of neuroscience and artificial intelligence technologies.
- 26 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 16 (4) , 3499-3511
DGC-Link: Dual-Gate Chebyshev Linkage Network on EEG Emotion Recognition
EEG emotion recognition presents several challenges, including region correlation, local and long-range node connectivity, and multi-channel patterns, necessitating advanced methods capable of effectively capturing and utilising complex EEG signal information. This paper introduces a novel method, the dual-gate Chebyshev Linkage network (DGC-Link), which comprises three main components: the Chebyshev Linkage (CL) module for extracting regional correlation features, the dual-gate module for regulating the flow of different-order information, and the deep network for extracting multi-channel features and enhancing representation capabilities. Validated on three datasets (SEED, DREAMER, and MPED) with ablation experiments demonstrating each component's effectiveness, DGC-Link achieves superior recognition performance compared to state-of-the-art methods. Notably, it achieves 96.43% accuracy on differential entropy on the SEED dataset, and 98.58%, 97.62%, and 98.01% for valence, arousal, and dominance classifications on the DREAMER dataset, along with 78.48% and 44.93% for 3-class and 7-class classifications on the MPED dataset. These results highlight DGC-Link's potential for improved performance in EEG emotion recognition.
- 26 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 16 (4) , 3499-3511
DGC-Link: Dual-Gate Chebyshev Linkage Network on EEG Emotion Recognition
EEG emotion recognition presents several challenges, including region correlation, local and long-range node connectivity, and multi-channel patterns, necessitating advanced methods capable of effectively capturing and utilising complex EEG signal information. This paper introduces a novel method, the dual-gate Chebyshev Linkage network (DGC-Link), which comprises three main components: the Chebyshev Linkage (CL) module for extracting regional correlation features, the dual-gate module for regulating the flow of different-order information, and the deep network for extracting multi-channel features and enhancing representation capabilities. Validated on three datasets (SEED, DREAMER, and MPED) with ablation experiments demonstrating each component's effectiveness, DGC-Link achieves superior recognition performance compared to state-of-the-art methods. Notably, it achieves 96.43% accuracy on differential entropy on the SEED dataset, and 98.58%, 97.62%, and 98.01% for valence, arousal, and dominance classifications on the DREAMER dataset, along with 78.48% and 44.93% for 3-class and 7-class classifications on the MPED dataset. These results highlight DGC-Link's potential for improved performance in EEG emotion recognition.
- 13 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (9) , 2924-2936
MF$^{2}$-Net: Exploring a Meta-Fuzzy Multimodal Fusion Network for Depression Recognition
Depression is a prevalent mental illness that significantly impacts the well-being of individuals and the development of society. The current diagnostic methods are largely subjective and time-consuming. Moreover, the existing machine learning-based depression recognition methods struggle to fully exploit the collaborative benefits between modalities, lack interpretability in their fusion processes, and perform inadequately in few-shot depression recognition tasks. To address these challenges, we propose a meta-fuzzy multimodal fusion network (MF$^{2}$-Net) for depression recognition. This innovative approach integrates physiological signals and behavioral data, employs multiple MLPs to learn the fuzzy measures of single base learners and complementary increments, then constructs all fuzzy measures, and finally achieves an interpretable decision-level fusion process through fuzzy integrals. Furthermore, we incorporate model-agnostic meta-learning for conducting few-shot domain-adaptive training, mitigating the issues related to high individual variability levels and the scarcity of multimodal depression data. Our method demonstrates exceptional classification performance in subject-independent experiments implemented on public datasets; offers a reliable solution for objectively, effectively, and conveniently recognizing depression; and has the potential to promote the clinical applications of rapid intelligent depression diagnosis.
- 13 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (9) , 2924-2936
MF$^{2}$-Net: Exploring a Meta-Fuzzy Multimodal Fusion Network for Depression Recognition
Depression is a prevalent mental illness that significantly impacts the well-being of individuals and the development of society. The current diagnostic methods are largely subjective and time-consuming. Moreover, the existing machine learning-based depression recognition methods struggle to fully exploit the collaborative benefits between modalities, lack interpretability in their fusion processes, and perform inadequately in few-shot depression recognition tasks. To address these challenges, we propose a meta-fuzzy multimodal fusion network (MF$^{2}$-Net) for depression recognition. This innovative approach integrates physiological signals and behavioral data, employs multiple MLPs to learn the fuzzy measures of single base learners and complementary increments, then constructs all fuzzy measures, and finally achieves an interpretable decision-level fusion process through fuzzy integrals. Furthermore, we incorporate model-agnostic meta-learning for conducting few-shot domain-adaptive training, mitigating the issues related to high individual variability levels and the scarcity of multimodal depression data. Our method demonstrates exceptional classification performance in subject-independent experiments implemented on public datasets; offers a reliable solution for objectively, effectively, and conveniently recognizing depression; and has the potential to promote the clinical applications of rapid intelligent depression diagnosis.
- 5 March 2025
- Published by Springer Nature
- p. 414-425
A Domain-Adapted Brain-Computer Interface for Implicit Emotion Tagging of Multimedia Content
- 5 March 2025
- Published by Springer Nature
- p. 414-425
A Domain-Adapted Brain-Computer Interface for Implicit Emotion Tagging of Multimedia Content