Performance of stochastic gradient descent adaptive beamforming using sonar data

Using sonar data characterised by the presence of a strong interference, adaptive beamforming is performed by the stochastic gradient descent algorithm in the frequency domain. The weights are constrained to have a unity response in the look direction. Convergence rates are estimated from averaged adaption curves, and, in most cases, these are found to be in reasonable agreement with theoretical values. It is conjectured that a marked discrepancy in one case is due to phase errors in the data covariance matrix.