An accurate perturbation analysis algorithm for music with Toeplitz covariance matrix

In this paper, a new perturbation analysis algorithm for the MUltiple SIgnal Classification (MUSIC) estimator applied to a Hermitian Toeplitz covariance matrix is presented. Inspired by the perspective that the MUSIC algorithm can be recognized as a structured matrix approximation, the perturbation of parameter estimates can be predicted more accurately by seeking the minimum of a Frobenius norm. The prediction results are analytically expressed through a weighted least squares method. The performance of the MUSIC estimators can also be predicted using our algorithm.