Radar classification of human gait abnormality based on sum-of-harmonics analysis

Radar technology for monitoring of human gait has recently gained interest in the fields of home security, medical diagnosis, assisted living and rehabilitation. Due to its remote, reliable and privacy-preserving sensing, radar is promising to become an effective tool for medical gait analysis. We show the influences of gait abnormalities and assistive walking devices on the joint-variable representations of the back-scattered radar signals. Using both, parametric and non-parametric techniques, we extract gait features to classify normal, abnormal and cane-assisted gait. In particular, the fundamental frequency of the time-frequency behavior is estimated using sum-of-harmonics modeling in order to characterize different gaits. Results obtained using experimental K-band radar data are presented for the person-specific and person-generic case.