Model-based classification of human motion: Particle filtering applied to the Micro-Doppler spectrum

In this article, a novel motion model-based particle filter implementation is proposed to classify human motion and to estimate key state variables, such as the motion type, i.e. running or walking, and the subject's height. Micro-Doppler spectrum is used as the observable information. The system and measurement models of the human movements are built using three parameters (relative torso velocity, height of the body, gait phase). The algorithm developed has been verified on simulated and experimental data.