A Computationally Light MUSIC Based Algorithm for Automotive RADARs

In this paper, a computationally light single-snapshot multiple signal classification (MUSIC) algorithm is presented for multidimensional estimation in the framework of automotive radar systems. In particular, for the sake of simplicity, we focus on the two-dimensional (2D) range-angle processing. The goal is to reduce the computational effort with respect to the standard 2D MUSIC implementation and to improve the resolution which can be obtained by classical Fourier transform processing. To this end, the decorrelation step, commonly achieved by standard 2D spatial smoothing strategy, is replaced by straight-way arranging the data to form a block Toeplitz matrix whose blocks are Toeplitz matrices. The corresponding 2D eigenspectrum is employed to derive the one-dimensional (1D) signal subspaces which are used to build two 1D MUSIC peseudospectra for estimating targets' ranges and angles, separately. The association problem is then resolved by computing the 2D MUSIC pseudospectrum only for a few search points arising from the combinations of the estimated ranges and angles. Numerical and experimental results show that the proposed algorithm improves the resolution (compared to the Fourier Transform) with a dramatic reduction of the computational burden (compared to the standard MUSIC).