Human Classification Based on Micro-Doppler Features of Walking Using Deep Learning Algorithm

In this paper, a human classification method based on the convolution neural network (CNN) is proposed. The proposed method firstly applies the short-time Fourier transform (STFT) to the radar received signal, which is denoised by the principal components analysis, to obtain the micro-Doppler spectrogram. Then, a new feature extraction scheme is proposed to extract 2-dimentional feature from the micro-Doppler spectrogram. Finally, we construct a CNN to realize the human classification. In order to validate the human classification method, training and testing data are produced by using a human walking model with linear frequency modulated continuous wave (LFMCW) and the additive white Gaussian noise is added to the simulation data. Simulations are carried out under different signal-to-noise ratio (SNR), and the trained neural network can distinct 5 people with up to 100% of accuracy with high SNR. Simulation results verify the superiority of the proposed method.