MUSIC-Based Joint DoA Estimation and Signal Enumeration

In this paper we describe a new criterion for the detection of the number of signals impinging on a M-element uniform linear array (ULA). Our criterion makes explicit use of the peak information of the MUSIC spectrum. Specifically, for each hypothesis of k sources, in addition to computing the noise variance estimate using the M-k smallest eigenvalues of the sample covariance matrix, the new criterion applies an additional correction term calculated from the k largest peaks of the MUSIC spectrum which is generated from the testing noise subspace of dimension M-k. We prove that the proposed criterion provides a consistent estimate of the number of signals and demonstrate that it has a better performance at low SNR for equal-power sources when compared with the original MDL-based signal number detection criterion [1]. Enumeration Ren Wu and Ioannis N. Psaromiligkos