Wideband source modeling and localization: a HOS-based approach

A polyspectra-based solution to the nonparametric multichannel system identification problem is proposed. As an application, the problems of wideband source bearing estimation an identification are considered. Simultaneous detection and parameter estimation of wideband sources arriving at a sensor array are performed using higher-order statistics. Source detection and consistent parameter estimation are approached via a polyspectra-based information theoretic criterion. The proposed methods, besides being insensitive to additive Gaussian noise, are not restricted to minimum phase equivalents and are asymptotically optimal for non-Gaussian signals with unknown distributions.<
>