Cumulant based approach for direction-of-arrival estimation of wideband sources

In array signal processing applications involving eigenstructure methods, direction finding algorithms are often derived by using the second order information of source signals. But in some cases the noise is not spatially white or the number of source signals is larger than that of the sensors, and the sources cannot be correctly resolved by using conventional algorithms exploiting second order statistics. A wideband source localization method utilizing higher order statistics (HOS) is presented. The information of signals of higher order greater than two is immune from additive Gaussian noise whether they are spatially, temporally white or colored, so that a priori knowledge on the noise statistics is unnecessary. In the algorithm presented the fourth order cumulant matrix is calculated in the transformed domain and a MUSIC-like estimator is used. However, this technique can be generalized for other eigenstructure methods such as ESPRIT and WSF.