A New Method for Fast Estimation of Number of Signals

Detection of the number of signals is a key issue in array signal processing. In the case of small number of snapshots, the performance of Akaike Information theoretic Criteria (AIC) and Maximum Description Length (MDL) will deteriorate due to the large estimation error of covariance matrix of observed data. In this paper a new method for detecting the number of signals based on Multi-Stage Wiener Filter (MSWF) is put forward, which only utilize the forward recursion of observed data and does not need the estimation and eigendecomposition of the covariance matrix, therefore is computationally advantageous. Finally numerical simulation shows that the proposed method can perform the fast estimation of the number of signals effectively, especially in the case of small number of snapshots.