A statistical analysis of state-space approaches to signal modeling

This paper addresses the problem of analyzing the perturbation of state-space approaches to signal modeling. Principal component techniques have been used in many algorithms involving frequency estimation and system identification. It is shown that in many instances the performance of these principal component based approaches is far better than that of classical methods. However, most of the performance evaluations of these methods are based on simulation results only. In this paper, the perturbation of the parameter estimate for a single sinusoid signal corrupted with noise is studied analytically using a power method