Adaptive Beamforming with the Transform Domain LMS (Least Mean-Square)

摘要

In adaptive digital filtering, the Transform Domain LMS algorithm (TRLMS) has shown the ability of improving the convergence rate of the Widrow-LMS algorithm at the expense of implementing a fixed orthogonal transform. In broadband beamforming, each sensor of an array is typically followed by a tapped delay line to provide frequency dependent array weighting. The weights can be adjusted adaptively to steer the nulls of the beampattern toward any undesired sources. For least Mean Square (LMS) based adaptive in a multisignal environment, convergence is highly data dependent and is characterized by highly disparate modes, resulting in slow and noisy adaptation. This work examines the feasibility of extending the TRLMS algorithm to the multichannel scenario of wideband beamforming for the purpose of improving convergence rates.