Automatic Extraction of Doppler Envelopes for Gait Analysis Using FMCW Radar: A Novel Approach for Spatio-Temporal Parameters Estimation

초록

This study presents a novel method for automatically extracting Doppler envelopes from Frequency-Modulated Continuous Wave (FMCW) radar signals for gait analysis. In contrast to conventional percentile-based approaches that require manual selection of Doppler envelopes for specific body parts (Spine, Leg, and Foot), the proposed contour-based method enables fully automated estimation of representative speed values from the Doppler map. Experiments were conducted on five participants with varying physical characteristics, and key gait parameters-such as walking speed, step length, and stride time-were estimated and compared against motion capture-based ground truth. The proposed method demonstrated relative errors typically below 10%, with key parameters such as Foot Speed and Leg Speed falling below the commonly cited 5% clinical threshold. Paired t-tests revealed no statistically significant differences between the proposed estimates and the ground truth across all gait parameters (p>0.05), supporting the method's accuracy and reliability.

키워드