Selected work

Representative Papers

Robust and Efficient Noisy-Label Time-Series Classification via Dynamic Time Warping Based Granular Ball Computing

Aug 12, 2026

This work addresses the susceptibility of dynamic time warping (DTW)-based nearest neighbor classifiers to label noise and their high computational cost during inference in time series classification. To mitigate these issues, the authors propose a granular-ball-based classification framework that aggregates temporally similar training samples into granular balls and performs classification at the granularity level. Two DTW-driven strategies for constructing granular balls are introduced. Experimental results on four benchmark datasets demonstrate that the proposed method significantly enhances robustness against symmetric label noise while substantially reducing the number of DTW computations required during inference. The approach thus achieves a favorable trade-off between classification accuracy, computational efficiency, and noise resilience.

0 citationsRead paper

Physics-Informed Synthetic Dataset and Denoising TIE-Reconstructed Phase Maps in Transient Flows Using Deep Learning

Apr 12, 2026

This work addresses the challenge of low-frequency artifacts in transport-of-intensity equation (TIE)-reconstructed phase maps for transient compressible flow imaging, which obscure critical features such as jets and shock waves, compounded by the absence of ground-truth paired data for denoising. The authors propose a zero-shot denoising method that requires no real labeled data: leveraging physics-informed priors, they procedurally synthesize flow structures consistent with fluid dynamics, then generate synthetic noisy–clean phase map pairs by combining forward TIE simulation with inverse Laplacian-based reconstruction to train a U-Net denoiser. This approach achieves, for the first time, zero-shot generalization of deep learning to non-repeatable transient flows, demonstrating a 13,260% improvement in signal-to-background ratio and a 100.8% enhancement in structural sharpness within jet regions on experimental data captured at 25,000 fps.

0 citationsRead paper

Wi-Fi Radar via Over-the-Air Referencing: Bridging Wi-Fi Sensing and Bistatic Radar

Feb 05, 2026

This work addresses the challenge of achieving phase-coherent radar sensing with asynchronous commercial Wi-Fi devices by proposing LoSRef, an over-the-air reference mechanism that leverages the line-of-sight (LoS) path as a self-calibration signal. This approach enables delay calibration and phase alignment without requiring wired connections or dedicated antennas. Building upon this method, the authors develop the first bistatic radar sensing framework deployable on off-the-shelf Wi-Fi hardware. By integrating channel impulse response extraction with delay-Doppler analysis, the system achieves physically interpretable sensing of sub-wavelength micromotions—such as human gait and respiration—and can even detect dynamic target signals up to 20 dB weaker than the dominant static multipath components.

0 citationsRead paper

Recent publications

Latest Papers

Robust and Efficient Noisy-Label Time-Series Classification via Dynamic Time Warping Based Granular Ball Computing

Aug 12, 2026

This work addresses the susceptibility of dynamic time warping (DTW)-based nearest neighbor classifiers to label noise and their high computational cost during inference in time series classification. To mitigate these issues, the authors propose a granular-ball-based classification framework that aggregates temporally similar training samples into granular balls and performs classification at the granularity level. Two DTW-driven strategies for constructing granular balls are introduced. Experimental results on four benchmark datasets demonstrate that the proposed method significantly enhances robustness against symmetric label noise while substantially reducing the number of DTW computations required during inference. The approach thus achieves a favorable trade-off between classification accuracy, computational efficiency, and noise resilience.

0 citationsRead paper

Physics-Informed Synthetic Dataset and Denoising TIE-Reconstructed Phase Maps in Transient Flows Using Deep Learning

Apr 12, 2026

This work addresses the challenge of low-frequency artifacts in transport-of-intensity equation (TIE)-reconstructed phase maps for transient compressible flow imaging, which obscure critical features such as jets and shock waves, compounded by the absence of ground-truth paired data for denoising. The authors propose a zero-shot denoising method that requires no real labeled data: leveraging physics-informed priors, they procedurally synthesize flow structures consistent with fluid dynamics, then generate synthetic noisy–clean phase map pairs by combining forward TIE simulation with inverse Laplacian-based reconstruction to train a U-Net denoiser. This approach achieves, for the first time, zero-shot generalization of deep learning to non-repeatable transient flows, demonstrating a 13,260% improvement in signal-to-background ratio and a 100.8% enhancement in structural sharpness within jet regions on experimental data captured at 25,000 fps.

0 citationsRead paper

Wi-Fi Radar via Over-the-Air Referencing: Bridging Wi-Fi Sensing and Bistatic Radar

Feb 05, 2026

This work addresses the challenge of achieving phase-coherent radar sensing with asynchronous commercial Wi-Fi devices by proposing LoSRef, an over-the-air reference mechanism that leverages the line-of-sight (LoS) path as a self-calibration signal. This approach enables delay calibration and phase alignment without requiring wired connections or dedicated antennas. Building upon this method, the authors develop the first bistatic radar sensing framework deployable on off-the-shelf Wi-Fi hardware. By integrating channel impulse response extraction with delay-Doppler analysis, the system achieves physically interpretable sensing of sub-wavelength micromotions—such as human gait and respiration—and can even detect dynamic target signals up to 20 dB weaker than the dominant static multipath components.

0 citationsRead paper