MUSIC with outlier rejection: a new tool for improving the estimation performance at the threshold region of the SNR

The root-mean square error (RMSE) of the MUSIC-based estimates of parameters of closely spaced sources increase dramatically as the signal-to-noise ratio (SNR) decreases helow the resolution threshold. To improve the threshold performance, a new parameter estimation technique is introduced. The technique is based on rejecting outliers by estimating the center and the width of signal interval. The simple algorithms developed are based on the properties of the MUSIC null spectrum and its derivative. The RMSE approximations and simulation results show that the proposed approach has superior estimation performance at the threshold region of the SNR.