Detection of signals in reverberation via beamspace adaptive matched filter

In this paper, the adaptive matched filter (AMF) detector is extended into beamspace via pre-processing beamformers to detect signals in shallow water reverberation. The number of beamformers is usually selected as less than that of array elements, so that dimension of the sample covariance matrix is reduced. With this reduction, less number of independent samples is required to form the sample covariance matrix. Diagonal loading technique is used to avoid the ill-conditioning problem of the estimated covariance matrix if we want to further reduce the number of samples. Performance of this beamspace diagonal loaded AMF (BS-DLAMF) detector was studied via computer simulation, and compared with AMF detector and Kelly's GLRT detector. The reverberation data used in computer simulations were produced using VENUS (a virtual underwater environment simulating software). Results from the simulation showed that the BS-DLAMF requires substantially less signal to reverberation ratio (SRR) than the full dimension AMF to get equivalent performance.