Performance of spatial smoothing algorithms for correlated sources

The problem of identifying the angles of arrival of a set of plane waves impinging on a narrowband array of sensors and related spectral analysis problems have been addressed with a large number of algorithms. One of the most popular techniques is the multiple signal classification (MUSIC). The major shortcoming of the MUSIC algorithm is that it performs poorly when the sources are highly correlated. Fortunately, two algorithms exist to overcome this problem-spatial smoothing (SS) and forward-backward spatial smoothing (FBSS). The performance of the SS technique depends on signal bearings and spatial separation. For the same smoothing, FBSS can offer improved performance, but this depends on the signal phases. Numerical results for the variance of the algorithms are given to illustrate the points made