Passive acoustic bearing estimation algorithms applied on hydro acoustic data

This thesis evaluates the bearing performance of three different bearing estimation algorithms in a real application, where the target is a moving surface vessel. The evaluated methods are Conventional Beamforming and two highresolution estimators Minimum Variance and MUSIC (MUltiple SIgnal Classification). The spatial covariance matrix is of great importance to obtain accurate bearing estimates in all three estimators, which means that it has to be well estimated. The condition of the spatial covariance matrix and the performance of the bearing estimators have been studied regarding the length of the estimation window (i.e. number of samples used for each bearing estimate) and recursive updating of the spatial covariance matrix. This has been performed by first detecting the target, and then tracking it during the entire recording. From measurements it has been shown that recursive updating of the spatial covariance matrix yields better bearing performance than using longer estimation windows. In addition, the computational time and the time delay is reduced. Both Conventional Beamforming and MUSIC yield a very accurate bearing performance, which is substantially better than the Minimum Variance estimator. The MUSIC method shows the best resolution properties and is able to separate two closely spaced sources.