Detection and classification of ground targets using a Doppler RADAR

Abstract

In this article, we address the problems of detection and classification of ground moving targets. We have applied different known Constant False Alarm Rate (CFAR) detectors in the Time-Frequency domain. The transformation to the frequency domain is done first by a simple short time Fourier transformation (STFT) then, due to the micro-Doppler signal properties, we have introduced the high resolution spectral method which is MUltiple SIgnal Classification (MUSIC). The studied detectors are: cell averaged (CA), SO, GO, OS and CMLD CFAR. With the second problem, we applied the hierarchical classification using five stages. We used in each stage a method of classification based either on GMM (Gaussian Mixture Model), MFCC (Mel-frequency Cepstral Coefficients), the dominant frequencies extracted by ESPRIT or the modulation of the micro-Doppler gait signature. The data base of μD signals is obtained from a Doppler radar and the obtained results are acceptable.