Logarithmic least mean square sound-field estimation

The need for high-resolution methods in PASSIVE SONAR is based on the finite aperture of the receiving array. Therefore, what these methods should try to do is to remove the effect of the spatial lowpass filter from the measured data. This paper tries to solve this "inverse" problem by using a logarithmic Least Mean Square parameter estimation technique. Here, it is essential to assume that the number of parameters to be estimated (sound field density) is much larger than the number of data available (hydrophones). Two implementations have been analysed: the steepest descent and the so-called WB iteration, which was developed some years ago at SACLANTCEN by Wagstaff and Berrou [1,2,3]. The results of these two iteration procedures have been compared against the conventional beamformer using perfectly known, and real data. Also, a statistical performance analysis of detection and resolution probability, and accuracy of bearing estimation has been carried out. Here, the comparison is made not only with the conventional beamformer but also with the Capon adaptive beamformer and the MUSIC orthogonal beamformer.