A High-Resolution Framework for Range-Doppler Frequency Estimation in Automotive Radar Systems

In this paper, a signal processing framework using high-resolution algorithms is applied to automotive radar systems for target separation purposes. In many critical use cases, targets sharing similar parameters, i.e., range and relative velocity, can hardly be discriminated in a 2-dimensional (2-D) range-Doppler spectrum via conventional radar signal processing techniques. The proposed framework by taking advantage of a 2-D model-based algorithm with high-resolution capability, i.e., modified matrix enhancement and matrix pencil (MMEMP), is able to separate the overlapping targets by jointly estimating their aforementioned target parameters in a precise manner. Additionally, to reduce computational cost, the presented framework could preselect the peaks (local maxima) of interest via a 2-D model order selection algorithm, i.e., 2-D subspace-based automatic model order selection (2-D SAMOS), and apply the MMEMP only in multi-target situations. Finally, simulation and experiments are carried out to evaluate the performance of the proposed high-resolution target separation framework.