Human Motion Analysis and Classification Using Radar Micro-Doppler Signatures

The ability to detect and analyze micro motions in human body is a crucial task in surveillance systems. Although video based systems are currently available to address this problem, but they need high computational resources and under good environmental lighting condition to capture high quality images. In this paper, a novel non-parametric method is presented to detect and calculate human gait speed while analyzing human micro motions based on radar micro-Doppler signatures to classify human motions. The analysis was applied to real data captured by 10 GHz radar from real human targets in a parking lot. Each individual was asked to perform different motions like walking, running, holding a bag while running, etc. The analysis of the gathered data revealed the human motion directions, number of steps taken per second, and whether the person is swinging arms while moving or not. Based on human motion structure and limitations, motion profile of each individual was recognizable to find the combinations between walking or running, and holding an object or swinging arms. We conclude that by adopting this method we can detect human motion profiles in radar based on micro motions of arms and legs in human body for surveillance applications in adverse weather conditions.