Simplification of the MUSIC algorithm using a neural network

The computation of the MUSIC null spectrum and the search for its minima constitute the most demanding computations for the MUSIC algorithm. The time delay entailed can limit the direction-finding performance under nonstationary conditions. In this paper a neural network that directly maps the noise eigenvectors into the spectral minima is described.