A State-Machine-Based Approach for Human Activity Classification Using MIMO Radar

This paper proposes a state-machine-based human activity recognition method for an indoor monitoring multiple-input multiple-output (MIMO) radar. Our method uses the transition of the Doppler radar-cross-section (RCS) in addition to the prior and posterior postures estimated by the radar to attain a high recognition rate. The experiment showed the proposed method achieves a 94.8% recognition rate on average.