Deep Learning Methods for Personnel Recognition based on Micro-Doppler Features

In this paper, we investigate the use of human gait micro-Doppler features for personnel recognition with a deep learning approach. Compared with conventional methods for radar-based human recognition, most existing schemes remain in discussing the distinction of different human motions. The proposed method employs a deep convolutional neural network (DCNN), which combines the inception parallel structure, to learn the necessary features of the time-frequency complex tensor that can be used to address the problem of recognizing different human subjects. Real data is collected relating to eight human subjects using a continuous wave radar operating at K-band. The experimental results illustrate that by using the human gait micro-Doppler features, the DCNN can achieve an accuracy rate of 96.9% on personnel recognition.