On statistical performance and computer simulation of super-resolution algorithm in signal processing

The MUSIC (MUltiple SIgnal Classification) algorithm has attracted considerable interest in the signal processing literature during the last two decades. This thesis investigated of the performance and possible improvements to the MUSIC algorithm, as applied to the resolution of two signals closely spaced in frequency. Simulations were used to study and analyze several aspects of the MUSIC algorithm. Results of these simulations illustrate the superiority of the MUSIC algorithm over the classical Fourier-based periodogram in resolving two closely spaced signals. The sensitivity of the MUSIC algorithm to the modification of several of its parameters was investigated in an attempt to find optimal parameter settings. The statistical efficiency of the MUSIC algorithm was studied by simulation method. Simulations were also used to investigate a proposed interleaving method for improving the MUSIC algorithm, proposed in 1997 by Halder and Kailath. The results of these simulations suggest that this modification to the MUSIC algorithm produce an improvement in both the statistical efficiency and performance.