Toward Unobtrusive In-Home Gait Analysis Based on Radar Micro-Doppler Signatures.

Area of Science:

• Engineering
• Biomedical Engineering
• Signal Processing

Background:

• Gait analysis is crucial for security, medical diagnosis, rehabilitation, and assisted living.
• Traditional gait monitoring methods can be invasive or lack long-term reproducibility.
• Radar offers a contactless sensing solution for gait analysis.

Purpose of the Study:

• To demonstrate the applicability of radar for gait classification.
• To develop and evaluate novel approaches for gait recognition using radar micro-Doppler signatures.
• To address the intra-motion category classification problem in gait recognition.

Main Methods:

• Utilized K-band radar to collect gait data from four subjects.
• Developed new classification approaches based on physical features, subspace features, and sum-of-harmonics modeling.
• Evaluated performance across five gait classes: normal, pathological, and assisted walks.

Main Results:

• Proposed radar-based gait classification methods outperformed existing techniques.
• Achieved an average classification accuracy of 93.8% across analyzed gait classes.
• Demonstrated a classification rate of 98.5% for a single gait class and ~80% for new individuals.

Conclusions:

• Radar micro-Doppler signatures effectively capture gait variations.
• Five distinct walking styles were recognized with high accuracy using radar.
• Radar-based gait sensing is a promising technology for contactless monitoring in healthcare and security.