A generalised sidelobe canceller architecture based on oversampled subband decompositions

Adaptive broadband beamforming can be performed in oversampled subband signals, whereby an independent beamformer is operated in each frequency band. This has been shown to result in a considerably reduced computational complexity. In this paper, we primarily investigate the convergence behaviour of the generalised sidelobe canceller (GSC) based on normalised least mean squares algorithm (NLMS) when operated in subbands. The minimum mean squared error can be limited, amongst other factors, by the aliasing present in the subbands. With regard to convergence speed, there is strong indication that the subband-GSC converges faster than a fullband counterpart of similar modelling capabilities. Simulations are presented.