Single Target Recognition Using a Low-Cost FMCW Radar Based on Spectrum Analysis

Radar Target classification has gained huge interest in many fields over the years in applications like road surveillance, smart gates, and autonomous driving. This is accompanied with the new advancements in the radar technologies that made radars a cheap optional replacement of cameras for such applications. Target recognition requires the extraction of features specific to each target. The process of feature extraction may differ depending on the type of radar. Range and Doppler profiles are two of the good features that can be used in that regard. In FMCW radars, although, the 2D FFT processing technique is used for the extraction of range and Doppler information, it also provides a Range-Doppler map that many use for the extraction of the Range and Doppler profiles. However, in low-Cost FMCW radars, these maps can be lacking in quality and resolution due to some limitations introduced by the hardware. In this paper, we propose a radar target classification method using a 24 GHz low-cost FMCW radar based on spectrum analysis. The features used in the classification are directly related to the range and Doppler profiles of the targets. The classification was done on four classes of targets which are pedestrians, cars, trucks, and motorcycles.