Signal preprocessing routines for the detection and classification of human micro-Doppler radar signatures

In this study, we introduce a radar sensor that we have developed for the micro-Doppler signature measurements together with three consecutive signal preprocessing routines that are to be used for primary detection and classification of human gait and movements. The developed C-band radar sensor has been used in a series of human walking and running experiments that were conducted at Mersin University facilities. First, the raw data collected in time domain are transformed to two-dimensional (2D) time-frequency plane to investigate the Doppler characteristics of human movements. To detect and extract the main body signature of the human, a Kalman filter-based torso signature detection methodology is introduced as the first preprocessing tool. Then, to be able to better detect the other body parts' signatures, the torso's response is thoroughly extracted from the 2D time-Doppler image as the second preprocessing routine by applying a circular shift procedure along each time-bin. As the last procedure, envelopes of the micro-Doppler signatures on the 2D time-Doppler image plane are determined by another routine based on human Doppler threshold method. To assess the success and the validity of the proposed pre-signal processing tools, they are applied to the measured data collected via different experiments of human movements. After applying all these steps, the torso signature and envelopes of all other body parts' signatures are marked and determined with good fidelity.