Human Activity Micro-Doppler Corner Feature Representation Method Based on Dual-Threshold Harris Extractor for Through-the-Wall Radar

Micro-Doppler corner signature is a subset of Doppler feature, which arises from the relative motion of human limb nodes. However, conventional micro-Doppler signature extraction methods have low precision on radar images with strong clutter and low resolution restrictions, which is difficult to meet the demand of through-the-wall (TWR) detection. To address these issues, this paper proposes a micro-Doppler corner representation method based on image segmentation and dual-threshold Harris corner extraction. Specifically, clutter suppression is first achieved using mean filter and Butterworth filter, resulting in distinctive TWR range and Doppler profiles. Then, the micro-Doppler corner feature image is evenly divided into segments. Furthermore, a corner extractor with two thresholds is performed on each segment. The smaller threshold is employed for detecting a larger number of corners, while the larger threshold enhances the detection of precise corners. Finally, pseudo corners are removed by SUSAN method. Experiments are conducted to verify the effectiveness of the proposed method.