Data modeling through robust nonlinear least squares

Abstract

A nonlinear least-squares (LS) method for modeling empirically obtained data in sensor array signal processing is developed. The new method is robust with respect to outliers and has a comparable computational complexity to the standard least-squares method. Robustness is achieved by introducing a nonlinear function which weights the squared error term in the LS criterion. Weighting functions for mixture of two Gaussian distributions are determined by maximum likelihood estimation theory. The strength of this algorithm is demonstrated by simulation examples for the direction-of-arrival (DOA) estimation problem. >