Yongkang Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yongkang Wang has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Jieyi Liu
Xidian University
2 shared publications - Maoguo Gong
Xidian University, Xi'an, China
2 shared publications - Yu Zhou
School of Electronic Engineering, Key Laboratory of Collaborative Intelligence Systems, Ministry of Education, Xidian University, Xian, Shaanxi, China
2 shared publications - Hao Li
Xidian University
1 shared publication - Hao Tian Shi
State Key Laboratory of Millimeter Waves School of Information Science and Engineering Southeast University Nanjing China
1 shared publication Jiaxi Guo
School of Electronic Engineering, Xidian University, Xian, China
1 shared publication
Latest Publications
- Open Access
- 11 August 2026
- Published by Wiley in Advanced Materials Technologies
- Vol. 11 (20) , e71241
Intelligent Programmable Metasurfaces for Robust Spatiotemporal 3D Human Pose Estimation
Programmable metasurfaces (PMs) provide a compelling platform for human sensing under conditions where vision‐based methods fail. However, conventional metasurface sensing relies exclusively on electromagnetic scattering data captured at the isolated and current time instant. Lacking integration with the historical temporal context, these single‐frame approaches suffer when severe indoor multipath fading distorts the transient scattering signatures, leading to estimation failures during deep fading events. To address this limitation, an intelligent PM system is proposed utilizing a multi‐frame spatiotemporal fusion approach, which treats human motion as a continuous evolution process. By exploiting the temporal dependencies across sequential frames through a Spatiotemporal Metasurface Pose Network (ST‐MetaPose), the system autonomously compensates for deeply faded instantaneous observation using kinematic context. For experimental validation, a prototype system featuring a 1‐bit PM consisting of a 30 × 30 array operating at 11.4 GHz is implemented. Experimental results demonstrate robust, continuous three‐dimensional 18‐keypoint pose estimation at 20 Hz, achieving a mean per‐joint position error of 0.078 m. This framework maintains high estimation stability even during destructive multipath fading, opening new avenues for intelligent electromagnetic surveillance and robust human‐machine interaction.
- 11 August 2026
- Published by Wiley in Advanced Materials Technologies
- Vol. 11 (20) , e71241
Open AccessIntelligent Programmable Metasurfaces for Robust Spatiotemporal 3D Human Pose Estimation
Programmable metasurfaces (PMs) provide a compelling platform for human sensing under conditions where vision‐based methods fail. However, conventional metasurface sensing relies exclusively on electromagnetic scattering data captured at the isolated and current time instant. Lacking integration with the historical temporal context, these single‐frame approaches suffer when severe indoor multipath fading distorts the transient scattering signatures, leading to estimation failures during deep fading events. To address this limitation, an intelligent PM system is proposed utilizing a multi‐frame spatiotemporal fusion approach, which treats human motion as a continuous evolution process. By exploiting the temporal dependencies across sequential frames through a Spatiotemporal Metasurface Pose Network (ST‐MetaPose), the system autonomously compensates for deeply faded instantaneous observation using kinematic context. For experimental validation, a prototype system featuring a 1‐bit PM consisting of a 30 × 30 array operating at 11.4 GHz is implemented. Experimental results demonstrate robust, continuous three‐dimensional 18‐keypoint pose estimation at 20 Hz, achieving a mean per‐joint position error of 0.078 m. This framework maintains high estimation stability even during destructive multipath fading, opening new avenues for intelligent electromagnetic surveillance and robust human‐machine interaction.
- 14 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (2) , 2287-2301
VIB-CTFN: A Multimodal Data Fusion Method Based on Variational Information Bottleneck Optimization for Radar Deceptive Jamming Discrimination
Existing radar anti-jamming methods are difficult to effectively integrate and utilize multimodal input data, resulting in limited jamming discrimination performance in complex deceptive jamming environments. In order to achieve more sufficient information utilization, VIB-CTFN, a convolutional neural network-Transformer fusion network based on variational information bottleneck optimization is proposed, which introduces an information-theoretic approach to solve the issue of information redundancy in current fusion strategies. The network leverages complementary feature extraction. The convolutional neural network is utilized to extract the microscopic electromagnetic features of signal modality, while Transformer captures the macroscopic long-range kinematic dependencies of data modality, and the multimodal representations are fused with the feature-level multi-head attention mechanism. During alternate training, variational information bottleneck compresses redundant and irrelevant information through variational approximation, optimizes the target critical feature representation, enhancing the generalization ability and robustness of the model. Simulation experiments show that the accuracy of VIB-CTFN in complex deceptive jamming scenarios reaches 99.79%, which is 9.47% and 3.33% higher than that of convolutional neural network and Transformer without variational information bottleneck optimization, while 18.48% and 16.41% higher than that of the traditional methods signal and data fusion-based traditional interference model, demonstrating that the method has a robust and efficient jamming discrimination capability in complex electromagnetic environments.
- 14 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (2) , 2287-2301
VIB-CTFN: A Multimodal Data Fusion Method Based on Variational Information Bottleneck Optimization for Radar Deceptive Jamming Discrimination
Existing radar anti-jamming methods are difficult to effectively integrate and utilize multimodal input data, resulting in limited jamming discrimination performance in complex deceptive jamming environments. In order to achieve more sufficient information utilization, VIB-CTFN, a convolutional neural network-Transformer fusion network based on variational information bottleneck optimization is proposed, which introduces an information-theoretic approach to solve the issue of information redundancy in current fusion strategies. The network leverages complementary feature extraction. The convolutional neural network is utilized to extract the microscopic electromagnetic features of signal modality, while Transformer captures the macroscopic long-range kinematic dependencies of data modality, and the multimodal representations are fused with the feature-level multi-head attention mechanism. During alternate training, variational information bottleneck compresses redundant and irrelevant information through variational approximation, optimizes the target critical feature representation, enhancing the generalization ability and robustness of the model. Simulation experiments show that the accuracy of VIB-CTFN in complex deceptive jamming scenarios reaches 99.79%, which is 9.47% and 3.33% higher than that of convolutional neural network and Transformer without variational information bottleneck optimization, while 18.48% and 16.41% higher than that of the traditional methods signal and data fusion-based traditional interference model, demonstrating that the method has a robust and efficient jamming discrimination capability in complex electromagnetic environments.
- 15 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (10) , 16647-16660
CLMFN: An Intelligent Discrimination Method of Deception Jamming Based on Multimodal Systems
The existing anti-jamming methods are based on single modality, lacking cross-modalities feature extraction and general discriminant algorithms. In order to increase the diversity of the discriminant scenarios, a multi-modal fusion model of jamming countermeasure is proposed. The model can collaboratively process cross-modalities information and prevent the loss of discrimination performance when lacking partial modalities, thus ensuring the all-round utilization of information. Specifically, a convolutional neural network and an attention-based bidirectional long short-term memory network model are constructed to extract complex envelope features of signal modality and kinematic time-series features of data modality, respectively. At the same time, the two features are fused to learn joint representations between cross modalities. The constraint condition is added to the loss function which can ensure various feature vectors are highly correlated, so that the discrimination remains good even facing the single modality. Extensive experimental results show that the performance of proposed multi-modal network is superior to the single modal network. Specifically, the proposed model exhibits robustness for different modalities and can intelligently adjust the dependence on different modalities.
- 15 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (10) , 16647-16660
CLMFN: An Intelligent Discrimination Method of Deception Jamming Based on Multimodal Systems
The existing anti-jamming methods are based on single modality, lacking cross-modalities feature extraction and general discriminant algorithms. In order to increase the diversity of the discriminant scenarios, a multi-modal fusion model of jamming countermeasure is proposed. The model can collaboratively process cross-modalities information and prevent the loss of discrimination performance when lacking partial modalities, thus ensuring the all-round utilization of information. Specifically, a convolutional neural network and an attention-based bidirectional long short-term memory network model are constructed to extract complex envelope features of signal modality and kinematic time-series features of data modality, respectively. At the same time, the two features are fused to learn joint representations between cross modalities. The constraint condition is added to the loss function which can ensure various feature vectors are highly correlated, so that the discrimination remains good even facing the single modality. Extensive experimental results show that the performance of proposed multi-modal network is superior to the single modal network. Specifically, the proposed model exhibits robustness for different modalities and can intelligently adjust the dependence on different modalities.