Human posture recognition based on multi-channel SAR at 77GHz

Human posture recognition is gaining increasing attention due to its widespread application. Besides optical and wearable methods for human posture recognition, radar is also an important sensor with the advantage of independence of the condition of work. In this paper, it is proposed to recognize human posture based on multi-channel Synthetic Aperture Radar (SAR) at 77GHz. We obtain 2D radar images of human postures through synthetic aperture technology, and then obtain 3D radar images through multi-channel interferometry technology. By analyzing radar images with different postures and different resolutions, we find that the height distribution characteristics of radar images are discrepant under different postures. We propose to use the statistical characteristics of their height distribution to discriminate common human postures, like standing, sitting/squatting and lying. The final experimental results verify the feasibility and effectiveness of proposed method.