Xingyu Zhu
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
Xingyu Zhu has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Xiangbo Shu
Nanjing University of Science and Technology
4 shared publications - Jinhui Tang
Nanjing Forestry University, Nanjing, China
2 shared publications - Binqian Xu
Nanjing University of Science and Technology
1 shared publication - Jiachao Zhang
Artificial Intelligence Industrial Technology Research Institute, Nanjing Institute of Technology, Nanjing, China
1 shared publication - Jisheng Kou
State Key Laboratory of Intelligent Deep Metal Mining and Equipment, Shaoxing University, Shaoxing 312000, Zhejiang, China
1 shared publication - Junjie Wang
Key Laboratory of Rock Mechanics and Geohazards of Zhejiang Province, Shaoxing 312000, P. R. China
1 shared publication
Latest Publications
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 1-12
Topology-Evolving Semantic Adaptation for Few-Shot Class-Incremental Action Recognition
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 1-12
Topology-Evolving Semantic Adaptation for Few-Shot Class-Incremental Action Recognition
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 27 (15209210) , 7053-7066
Prompt-Guided Prototype-Aware Commonality and Discrimination Learning for Zero-Shot Skeleton-Based Action Recognition
Zero-Shot Skeleton-Based Action Recognition (ZSSAR) is an emerging research field focused on developing alignment models that connect skeleton movements with action definitions, thus enabling generalization to unobserved actions. Current methods often employ generative models to reconstruct cross-modal features or enhance mutual information across modalities for alignment. However, when applied to unseen action categories, these models often neglect the inherent consistency among basic actions, thereby diminishing their generalization capabilities. Furthermore, imprecise annotations fail to capture the rich semantic details of actions, resulting in misalignment. Inspired by human cognitive processes and chain of thought, we argue that integrating prior information about human actions with intrinsic commonality knowledge of basic actions is essential for ZSSAR. To actualize this, we propose a novel method termed Prompt-guided Prototype-aware Commonality and Discrimination Learning (PP-CDL). This method utilize the comprehensive world knowledge contained in LLMs, employing tailored prompts to partition seen action categories into distinct, non-overlapping prototype spaces that embody the commonality knowledge of basic actions. Subsequently, we introduce the Inter- and Intra-Prototype Discriminating (I2PD) module and the Intra-Prototype Commonality Mining (IPCM) module. The I2PD amplifies the distinctiveness of knowledge within prototypes, furnishing a personalized search space for the recognition of unseen actions. In contrast, the IPCM models the shared commonality concept within prototypes, bolstering the consistency between skeleton action representations and corresponding text knowledge representations. Experiments on different skeleton action benchmarks demonstrate the significant improvement of our method over existing alternatives.
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 27 (15209210) , 7053-7066
Prompt-Guided Prototype-Aware Commonality and Discrimination Learning for Zero-Shot Skeleton-Based Action Recognition
Zero-Shot Skeleton-Based Action Recognition (ZSSAR) is an emerging research field focused on developing alignment models that connect skeleton movements with action definitions, thus enabling generalization to unobserved actions. Current methods often employ generative models to reconstruct cross-modal features or enhance mutual information across modalities for alignment. However, when applied to unseen action categories, these models often neglect the inherent consistency among basic actions, thereby diminishing their generalization capabilities. Furthermore, imprecise annotations fail to capture the rich semantic details of actions, resulting in misalignment. Inspired by human cognitive processes and chain of thought, we argue that integrating prior information about human actions with intrinsic commonality knowledge of basic actions is essential for ZSSAR. To actualize this, we propose a novel method termed Prompt-guided Prototype-aware Commonality and Discrimination Learning (PP-CDL). This method utilize the comprehensive world knowledge contained in LLMs, employing tailored prompts to partition seen action categories into distinct, non-overlapping prototype spaces that embody the commonality knowledge of basic actions. Subsequently, we introduce the Inter- and Intra-Prototype Discriminating (I2PD) module and the Intra-Prototype Commonality Mining (IPCM) module. The I2PD amplifies the distinctiveness of knowledge within prototypes, furnishing a personalized search space for the recognition of unseen actions. In contrast, the IPCM models the shared commonality concept within prototypes, bolstering the consistency between skeleton action representations and corresponding text knowledge representations. Experiments on different skeleton action benchmarks demonstrate the significant improvement of our method over existing alternatives.
- 14 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 34 (10577149) , 4908-4922
Client-Unbiased Skeletal Action Recognizer in Federated Learning
Edge sensor devices generate vast amounts of user data, but centralized processing poses privacy risks. Federated Learning addresses this by decentralizing training. However, applying Federated Learning directly to skeleton videos fails to preserve motion dynamics and suffers from client heterogeneity bias. To address these limitations, we propose CSAR—a Client-Unbiased Skeletal Action Recognizer for Federated Learning— which tackles two core challenges: motion dynamics preservation and classifier bias mitigation. Specifically, CSAR employs a Model Calibration Loss during client training to align client-server representations and reduce drift. On the server, it generates class-balanced spatiotemporal federated features through Prototypical Gaussian Sampling, subsequently refined via a Motion-aware Differential Loss to capture kinematic properties. These features enable retraining of a globally debiased recognizer that achieves accuracy comparable to real-data-trained models. Further stabilization is achieved through Knowledge Matching, which enhances global understanding. Experiments under natural and label heterogeneity confirm that CSAR outperforms state-of-the-art methods.
- 14 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 34 (10577149) , 4908-4922
Client-Unbiased Skeletal Action Recognizer in Federated Learning
Edge sensor devices generate vast amounts of user data, but centralized processing poses privacy risks. Federated Learning addresses this by decentralizing training. However, applying Federated Learning directly to skeleton videos fails to preserve motion dynamics and suffers from client heterogeneity bias. To address these limitations, we propose CSAR—a Client-Unbiased Skeletal Action Recognizer for Federated Learning— which tackles two core challenges: motion dynamics preservation and classifier bias mitigation. Specifically, CSAR employs a Model Calibration Loss during client training to align client-server representations and reduce drift. On the server, it generates class-balanced spatiotemporal federated features through Prototypical Gaussian Sampling, subsequently refined via a Motion-aware Differential Loss to capture kinematic properties. These features enable retraining of a globally debiased recognizer that achieves accuracy comparable to real-data-trained models. Further stabilization is achieved through Knowledge Matching, which enhances global understanding. Experiments under natural and label heterogeneity confirm that CSAR outperforms state-of-the-art methods.
- Open Access
- 14 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 258-263
Language-Guided Test-time Prompt Tuning for Zero-shot Skeleton-based Action Recognition
- 14 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 258-263
Open AccessLanguage-Guided Test-time Prompt Tuning for Zero-shot Skeleton-based Action Recognition
- Open Access
- 21 January 2021
- Published by Yandy Scientific Press in Capillarity
- Vol. 4 (1) , 1-12
Water uptake in parallel fractures
Water uptake in rock fractures caused by rainfall plays a significant role in slope stability analysis. Since the fracture network system has complicated structures and multiple scales, the models based on the averaged system cannot account for these properties. On the other hand, a model describing a single fracture with fractal characteristics and surface roughness fails to deal with the case of multiple fractures at spatial scales. In this study, a fracture-network model is established to account for the complex structures and multiple scales of fractures. By considering the connectivity between fractures and the limited area of aquifer, capillary pressure formulations in different fractures are derived based on the Young-Laplace equation, and the final water level under specific rainfall conditions is also obtained. The cross-section shapes and exhaust conditions of rainwater infiltration have important influences on the final water level. The results indicate that the final water level is proportional to the ratio of perimeter to cross-section area when the fracture is a cylinder, and a circular pipe can reduce water level elevation in the fracture system. Cited as: Wang, J., Zhu, X., Pan, Y., Kou, J., Sun, S. Water uptake in parallel fractures. Capillarity, 2021, 4(1): 1-12, doi: 10.46690/capi.2021.01.01
- 21 January 2021
- Published by Yandy Scientific Press in Capillarity
- Vol. 4 (1) , 1-12
Open AccessWater uptake in parallel fractures
Water uptake in rock fractures caused by rainfall plays a significant role in slope stability analysis. Since the fracture network system has complicated structures and multiple scales, the models based on the averaged system cannot account for these properties. On the other hand, a model describing a single fracture with fractal characteristics and surface roughness fails to deal with the case of multiple fractures at spatial scales. In this study, a fracture-network model is established to account for the complex structures and multiple scales of fractures. By considering the connectivity between fractures and the limited area of aquifer, capillary pressure formulations in different fractures are derived based on the Young-Laplace equation, and the final water level under specific rainfall conditions is also obtained. The cross-section shapes and exhaust conditions of rainwater infiltration have important influences on the final water level. The results indicate that the final water level is proportional to the ratio of perimeter to cross-section area when the fracture is a cylinder, and a circular pipe can reduce water level elevation in the fracture system. Cited as: Wang, J., Zhu, X., Pan, Y., Kou, J., Sun, S. Water uptake in parallel fractures. Capillarity, 2021, 4(1): 1-12, doi: 10.46690/capi.2021.01.01
- 1 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Robust Adaptive Beamforming for Coprime Array Based on Interference Power Estimation via Subspace
Compared with uniform linear array (ULA), the coprime array can obtain larger array aperture with the same number of physical sensors, which leads to higher spatial resolution and less effect of mutual coupling. Also, the degrees of freedom (DOFs) can be increased by reconstructing the sampling covariance matrix (SCM) in virtual array domain. In this paper, we propose a robust adaptive beamforming (RAB) algorithm based on coprime array. A novel subspace based method is performed in virtual array domain to estimate the power and direction of arrival (DOA) of interference, which also obtains the ability for DOFs enhancement. Simulation results demonstrate the robustness and effectiveness of the proposed method.
- 1 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Robust Adaptive Beamforming for Coprime Array Based on Interference Power Estimation via Subspace
Compared with uniform linear array (ULA), the coprime array can obtain larger array aperture with the same number of physical sensors, which leads to higher spatial resolution and less effect of mutual coupling. Also, the degrees of freedom (DOFs) can be increased by reconstructing the sampling covariance matrix (SCM) in virtual array domain. In this paper, we propose a robust adaptive beamforming (RAB) algorithm based on coprime array. A novel subspace based method is performed in virtual array domain to estimate the power and direction of arrival (DOA) of interference, which also obtains the ability for DOFs enhancement. Simulation results demonstrate the robustness and effectiveness of the proposed method.