Yang Li
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
Yang Li has a total of 79 co-authors. Here are the top co-authors by number of shared publications:
- Bin Zhao
Research Institute of Electronic Engineering, Harbin Institute of Technology, Harbin, China
14 shared publications - Xiao Xu
Research Institute of Electronic Engineering, Harbin Institute of Technology, Harbin, China
13 shared publications - Fanglin Chen
Research Institute of Electronic Engineering, Harbin Institute of Technology, Harbin, China
11 shared publications - Chunmao Yeh
Beijing Institute of Radio Measurement, Beijing, China
10 shared publications - Yunrong Zhu
Research Institute of Electronic Engineering, Harbin Institute of Technology, Harbin, China
10 shared publications - Wenbo Ding
Correspondence to: Dr. Hongfa Zhao, Prof. Wenbo Ding, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, Guangdong, China.
9 shared publications
Latest Publications
- 20 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 11619-11635
UDPE: Unsupervised Dual-Parameter Estimation for Radar Target Detection in HFSWR With Weibull Clutter
The performance of existing moment-based constant false alarm rate (CFAR) detectors is constrained by the limited availability of homogeneous samples and the difficulty in modeling the local correlation of clutter, which result in inaccurate estimation of clutter distribution parameters. Although neural networks (NNs) have the potential to exploit the spatial correlations among clutter samples to enhance estimation accuracy, their deployment in practical radar scenarios remains challenging due to the unavailability of true clutter parameters, the inconsistency of estimation errors, and the inherent randomness of NNs. To overcome these challenges, an unsupervised dual-parameter estimation (UDPE) method within an ensemble learning framework is proposed for target detection in high-frequency surface wave radar (HFSWR). First, to overcome the problem of unavailable ground truth, an unsupervised moment estimation network is developed, which is inspired by the concept of data reconstruction and accurately estimates the three moments of clutter data by utilizing the samples surrounding the cell under test (CUT). Second, we identify and theoretically analyze that the inconsistency of NN moment estimation errors significantly affects the accuracy of shape parameter estimation. To alleviate this issue, a multi-channel moment estimation network is constructed to exploit inter-channel correlations and improve error consistency. Third, to mitigate the problem of random fluctuations inherent in single-network moment estimation, an ensemble unsupervised moment estimation network (EUME-Net) is designed, which leverages model aggregation to reduce variance and enhance overall estimation accuracy. Finally, experiments on both simulated and measured shore-based HFSWR data demonstrate that the proposed method achieves superior target detection performance through more accurate and efficient parameter estimation.
- 20 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 11619-11635
UDPE: Unsupervised Dual-Parameter Estimation for Radar Target Detection in HFSWR With Weibull Clutter
The performance of existing moment-based constant false alarm rate (CFAR) detectors is constrained by the limited availability of homogeneous samples and the difficulty in modeling the local correlation of clutter, which result in inaccurate estimation of clutter distribution parameters. Although neural networks (NNs) have the potential to exploit the spatial correlations among clutter samples to enhance estimation accuracy, their deployment in practical radar scenarios remains challenging due to the unavailability of true clutter parameters, the inconsistency of estimation errors, and the inherent randomness of NNs. To overcome these challenges, an unsupervised dual-parameter estimation (UDPE) method within an ensemble learning framework is proposed for target detection in high-frequency surface wave radar (HFSWR). First, to overcome the problem of unavailable ground truth, an unsupervised moment estimation network is developed, which is inspired by the concept of data reconstruction and accurately estimates the three moments of clutter data by utilizing the samples surrounding the cell under test (CUT). Second, we identify and theoretically analyze that the inconsistency of NN moment estimation errors significantly affects the accuracy of shape parameter estimation. To alleviate this issue, a multi-channel moment estimation network is constructed to exploit inter-channel correlations and improve error consistency. Third, to mitigate the problem of random fluctuations inherent in single-network moment estimation, an ensemble unsupervised moment estimation network (EUME-Net) is designed, which leverages model aggregation to reduce variance and enhance overall estimation accuracy. Finally, experiments on both simulated and measured shore-based HFSWR data demonstrate that the proposed method achieves superior target detection performance through more accurate and efficient parameter estimation.
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 9892-9907
Conditional Maximum Likelihood Estimation and Cramér–Rao Lower Bound for High-Speed Multitarget Motion Parameter Estimation in Homogeneous Cluttered Environments
This paper presents a focused study on multi-target motion parameter estimation under homogeneous clutter. We first derive an exact maximum likelihood estimation (MLE) formulation that extends the conventional Radon–Fourier transform (RFT) framework to handle multiple targets and clutter. To reduce the high computational load of joint multi-target MLE, we introduce a con ditional MLE (C-MLE) framework that decouples the parameters. In C-MLE, weestimate only one target parameter at each step while keeping the other parameters fixed. We then update the parameters of all targets in turn. Theoretical analysis and experiments show that, compared with existing methods, C-MLE can remove the interference from known targets and improve the estimation of the remaining targets. In addition, we derive the Cram´ er–Rao lower bound (CRLB) as a benchmark for performance evaluation. Finally, simulation results show that the proposed C-MLE achieves lower mean squared error (MSE) and approaches the CRLB in high signal-to-clutter-plus-noise ratio (SCNR) regime, which confirms its strong performance in multi-target cluttered scenarios.
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 9892-9907
Conditional Maximum Likelihood Estimation and Cramér–Rao Lower Bound for High-Speed Multitarget Motion Parameter Estimation in Homogeneous Cluttered Environments
This paper presents a focused study on multi-target motion parameter estimation under homogeneous clutter. We first derive an exact maximum likelihood estimation (MLE) formulation that extends the conventional Radon–Fourier transform (RFT) framework to handle multiple targets and clutter. To reduce the high computational load of joint multi-target MLE, we introduce a con ditional MLE (C-MLE) framework that decouples the parameters. In C-MLE, weestimate only one target parameter at each step while keeping the other parameters fixed. We then update the parameters of all targets in turn. Theoretical analysis and experiments show that, compared with existing methods, C-MLE can remove the interference from known targets and improve the estimation of the remaining targets. In addition, we derive the Cram´ er–Rao lower bound (CRLB) as a benchmark for performance evaluation. Finally, simulation results show that the proposed C-MLE achieves lower mean squared error (MSE) and approaches the CRLB in high signal-to-clutter-plus-noise ratio (SCNR) regime, which confirms its strong performance in multi-target cluttered scenarios.
- 23 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 9690-9705
Two-Stage Coherent Integration for High-Speed Maneuvering Target in Wideband Radar Based on FA/SA-RPIT
High-speed maneuvering targets in wideband radar present severe challenges for coherent integration due to range migration (RM), Doppler frequency migration (DFM), and range spread (RS). To address these issues, this paper proposes the Full Aperture and Sub-Aperture Range-Phase-Invariant Transforms (FA-RPIT and SA-RPIT), which jointly mitigate energy dispersion and enable coherent accumulation across multiple scatterers. The proposed transforms achieve invariance to both the initial range and phase of scatterers, allowing effective envelope and phase alignment. Theoretical analysis and simulations demonstrate an output SNR gain of at least 10log10 K dB as the number of scatterers K increases, while both simulated and measured data confirm the superior performance and practical effectiveness of the proposed methods in wideband radar scenarios.
- 23 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 9690-9705
Two-Stage Coherent Integration for High-Speed Maneuvering Target in Wideband Radar Based on FA/SA-RPIT
High-speed maneuvering targets in wideband radar present severe challenges for coherent integration due to range migration (RM), Doppler frequency migration (DFM), and range spread (RS). To address these issues, this paper proposes the Full Aperture and Sub-Aperture Range-Phase-Invariant Transforms (FA-RPIT and SA-RPIT), which jointly mitigate energy dispersion and enable coherent accumulation across multiple scatterers. The proposed transforms achieve invariance to both the initial range and phase of scatterers, allowing effective envelope and phase alignment. Theoretical analysis and simulations demonstrate an output SNR gain of at least 10log10 K dB as the number of scatterers K increases, while both simulated and measured data confirm the superior performance and practical effectiveness of the proposed methods in wideband radar scenarios.
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (9) , 19629-19648
WiC-RadGen: Dual-Band Wi-Fi Conditioned Radar Point Cloud Generation via Discrete VAEs and Masked Autoregressive Transformers
Wireless sensing technologies are crucial for a wide range of real-world applications, from indoor navigation and smart home automation to autonomous vehicle perception and industrial monitoring. While dedicated millimeter-wave (mmWave) radar systems offer high-precision spatial measurements, they encounter deployment challenges due to their high cost and power consumption. In contrast, Wi-Fi-based sensing leverages existing communication infrastructure, offering cost-effective alternatives via channel state information (CSI) in 2.4-GHz systems and beamforming in mmWave Wi-Fi (60 GHz). However, traditional Wi-Fi sensing provides lower spatial fidelity than radar systems. This study introduces Wi-Fi Conditioned Radar point cloud Generation (WiC-RadGen), which closes this performance gap by generating radar-quality point clouds from dual-band Wi-Fi signals. Our approach employs a two-stage architecture. Initially, a discrete variational autoencoder (DVAE) with specialized loss functions (Chamfer distance (CD), annealed Kullback–Leibler (KL) divergence, and cluster loss) tokenizes sparse radar point clouds while preserving geometric coherence. Subsequently, a conditional masked autoregressive transformer (CMAR) integrates 2.4-GHz CSI and 60-GHz beam signal-to-noise ratio (SNR) through cross-attention mechanisms for sequential token prediction. Technical innovations include an empty-token predictor (ETP) with a differential Neyman–Pearson (DNP) loss that efficiently identifies uninformative regions, focal loss to address class imbalance, and iterative unmasking for high-fidelity generation. Experimental validation demonstrates competitive localization accuracy approaching mmWave radar performance while retaining Wi-Fi’s deployment advantages. WiC-RadGen advances wireless sensing by enabling radar-precision applications through ubiquitous Wi-Fi infrastructure, reducing deployment costs without compromising spatial fidelity.
- 12 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (9) , 19629-19648
WiC-RadGen: Dual-Band Wi-Fi Conditioned Radar Point Cloud Generation via Discrete VAEs and Masked Autoregressive Transformers
Wireless sensing technologies are crucial for a wide range of real-world applications, from indoor navigation and smart home automation to autonomous vehicle perception and industrial monitoring. While dedicated millimeter-wave (mmWave) radar systems offer high-precision spatial measurements, they encounter deployment challenges due to their high cost and power consumption. In contrast, Wi-Fi-based sensing leverages existing communication infrastructure, offering cost-effective alternatives via channel state information (CSI) in 2.4-GHz systems and beamforming in mmWave Wi-Fi (60 GHz). However, traditional Wi-Fi sensing provides lower spatial fidelity than radar systems. This study introduces Wi-Fi Conditioned Radar point cloud Generation (WiC-RadGen), which closes this performance gap by generating radar-quality point clouds from dual-band Wi-Fi signals. Our approach employs a two-stage architecture. Initially, a discrete variational autoencoder (DVAE) with specialized loss functions (Chamfer distance (CD), annealed Kullback–Leibler (KL) divergence, and cluster loss) tokenizes sparse radar point clouds while preserving geometric coherence. Subsequently, a conditional masked autoregressive transformer (CMAR) integrates 2.4-GHz CSI and 60-GHz beam signal-to-noise ratio (SNR) through cross-attention mechanisms for sequential token prediction. Technical innovations include an empty-token predictor (ETP) with a differential Neyman–Pearson (DNP) loss that efficiently identifies uninformative regions, focal loss to address class imbalance, and iterative unmasking for high-fidelity generation. Experimental validation demonstrates competitive localization accuracy approaching mmWave radar performance while retaining Wi-Fi’s deployment advantages. WiC-RadGen advances wireless sensing by enabling radar-precision applications through ubiquitous Wi-Fi infrastructure, reducing deployment costs without compromising spatial fidelity.
- Open Access
- 14 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19  (19391404) , 8250-8262
A Hybrid Architecture Multistatic Track Initiation Method Based on Spatiotemporal Feature Fusion Network
This paper proposes an intelligent track initiation method based on spatio-temporal feature fusion network for multistatic radar systems, addressing the challenge of multi-target track initiation in complex electromagnetic environments. The technique leverages a combination of a Spatio-Temporal Feature-driven Threshold Network (STFTN) and a Spatio-Temporal Graph Convolutional Network (STGCN) within a hybrid architecture, combining the computational efficiency of distributed processing with the information richness of centralized processing. Initially, a coarse screening is performed locally at the sub-processor using the speed limit method. Subsequently, relevant data is sent to a central processor where the STFTN dynamically generates intelligent thresholds to optimize the correlation range for candidate tracks. Finally, motion features of candidate tracks are extracted and fed into the STGCN in the central processor for secondary intelligent judgment, effectively distinguishing true tracks from false alarms. Simulation results demonstrate that the proposed algorithm achieves higher track initiation rates and fewer false tracks compared with traditional methods.
- 14 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19  (19391404) , 8250-8262
Open AccessA Hybrid Architecture Multistatic Track Initiation Method Based on Spatiotemporal Feature Fusion Network
This paper proposes an intelligent track initiation method based on spatio-temporal feature fusion network for multistatic radar systems, addressing the challenge of multi-target track initiation in complex electromagnetic environments. The technique leverages a combination of a Spatio-Temporal Feature-driven Threshold Network (STFTN) and a Spatio-Temporal Graph Convolutional Network (STGCN) within a hybrid architecture, combining the computational efficiency of distributed processing with the information richness of centralized processing. Initially, a coarse screening is performed locally at the sub-processor using the speed limit method. Subsequently, relevant data is sent to a central processor where the STFTN dynamically generates intelligent thresholds to optimize the correlation range for candidate tracks. Finally, motion features of candidate tracks are extracted and fed into the STGCN in the central processor for secondary intelligent judgment, effectively distinguishing true tracks from false alarms. Simulation results demonstrate that the proposed algorithm achieves higher track initiation rates and fewer false tracks compared with traditional methods.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64, 5011513
Radar-Referenced Prior Network for Infrared Iceberg Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64, 5011513
Radar-Referenced Prior Network for Infrared Iceberg Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62, 16455-16472
Prior Knowledge-Aided Contrastive Learning for HFSWR Target Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62, 16455-16472
Prior Knowledge-Aided Contrastive Learning for HFSWR Target Detection
- 6 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 1309-1324
A Neural Network CFAR Detector Based on Local Clutter Power Estimation Under Unsupervised Learning for HFSWR
High-frequency surface wave radar (HFSWR) has been widely used for ocean monitoring in recent years. However, its target detection faces challenges in complex non-homogeneous environments. Constant false alarm rate (CFAR) detectors degrade in performance due to inaccurate local clutter power estimation (LCPE), while the performance of data-driven intelligent radar target detectors is limited by the challenges of data annotation and poor generalization. To address the above problems in single-pulse processing, this paper integrates CFAR and deep learning theories, and presents a neural network (NN) CFAR detector based on LCPE under unsupervised learning for HFSWR (ULCPE-CFAR). First, by analyzing the relationship between LCPE and CFAR in different environments, we clarify the impact of LCPE accuracy on detection performance, and then theoretically validate the feasibility of NN-based LCPE. Second, based on the above research ideas, we propose the ULCPE-CFAR. The network of ULCPE-CFAR is trained through self-supervised learning on clutter data reconstruction, followed by fine-tuning the optimal model to enhance detection performance in unknown data. During inference, ULCPE-CFAR employs the group masking strategy to mitigate background information loss, ensuring accurate LCPE. Experimental results demonstrate that the proposed ULCPE-CFAR exhibits superior performance compared to traditional detectors in complex environments, providing a reliable and practical solution for unsupervised intelligent radar detection.
- 6 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62  (00189251) , 1309-1324
A Neural Network CFAR Detector Based on Local Clutter Power Estimation Under Unsupervised Learning for HFSWR
High-frequency surface wave radar (HFSWR) has been widely used for ocean monitoring in recent years. However, its target detection faces challenges in complex non-homogeneous environments. Constant false alarm rate (CFAR) detectors degrade in performance due to inaccurate local clutter power estimation (LCPE), while the performance of data-driven intelligent radar target detectors is limited by the challenges of data annotation and poor generalization. To address the above problems in single-pulse processing, this paper integrates CFAR and deep learning theories, and presents a neural network (NN) CFAR detector based on LCPE under unsupervised learning for HFSWR (ULCPE-CFAR). First, by analyzing the relationship between LCPE and CFAR in different environments, we clarify the impact of LCPE accuracy on detection performance, and then theoretically validate the feasibility of NN-based LCPE. Second, based on the above research ideas, we propose the ULCPE-CFAR. The network of ULCPE-CFAR is trained through self-supervised learning on clutter data reconstruction, followed by fine-tuning the optimal model to enhance detection performance in unknown data. During inference, ULCPE-CFAR employs the group masking strategy to mitigate background information loss, ensuring accurate LCPE. Experimental results demonstrate that the proposed ULCPE-CFAR exhibits superior performance compared to traditional detectors in complex environments, providing a reliable and practical solution for unsupervised intelligent radar detection.
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25  (17) , 33100-33114
RM-GMSD and FRM-GMSD: Subspace-Based Detection of High-Speed Range-Spread Targets in Cluttered Environments
This paper addresses high-speed range-spread target detection in cluttered environments, considering both range migration (RM) and range-spread (RS) effects, as well as the disturbance of clutter. The target is modeled as multiple scatters with identical velocities. We first propose the Generalized Matched Subspace Detector for RM targets (RM-GMSD), which projects pulse-compressed echo data onto a joint subspace of multiple hypothesized scatter echoes. However, RM-GMSD lacks compatibility with conventional extended target detection frameworks, preventing direct application of effective scatter selection algorithms, which are crucial for mitigating collapse loss caused by weak or ineffective scatters. To overcome this, we analyze RM-GMSD’s geometric projection properties and propose the Flexible RM-GMSD (FRM-GMSD). Unlike RM-GMSD, FRM-GMSD exploits the orthogonality of subspaces associated with different hypothesized scatters across range cells. It follows a two-stage process: first, projecting echo data onto multiple one-dimensional subspaces, then performing noncoherent integration. This reformulation aligns FRM-GMSD with conventional RS target detection framework, enabling effective scatter selection. Notably, incorporating effective scatter information allows FRM-GMSD to achieve optimal detection performance. Finally, numerical simulations and real-world data validate the effectiveness of the proposed approach.
- 21 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25  (17) , 33100-33114
RM-GMSD and FRM-GMSD: Subspace-Based Detection of High-Speed Range-Spread Targets in Cluttered Environments
This paper addresses high-speed range-spread target detection in cluttered environments, considering both range migration (RM) and range-spread (RS) effects, as well as the disturbance of clutter. The target is modeled as multiple scatters with identical velocities. We first propose the Generalized Matched Subspace Detector for RM targets (RM-GMSD), which projects pulse-compressed echo data onto a joint subspace of multiple hypothesized scatter echoes. However, RM-GMSD lacks compatibility with conventional extended target detection frameworks, preventing direct application of effective scatter selection algorithms, which are crucial for mitigating collapse loss caused by weak or ineffective scatters. To overcome this, we analyze RM-GMSD’s geometric projection properties and propose the Flexible RM-GMSD (FRM-GMSD). Unlike RM-GMSD, FRM-GMSD exploits the orthogonality of subspaces associated with different hypothesized scatters across range cells. It follows a two-stage process: first, projecting echo data onto multiple one-dimensional subspaces, then performing noncoherent integration. This reformulation aligns FRM-GMSD with conventional RS target detection framework, enabling effective scatter selection. Notably, incorporating effective scatter information allows FRM-GMSD to achieve optimal detection performance. Finally, numerical simulations and real-world data validate the effectiveness of the proposed approach.