Adaptive Wideband Beamforming with Combined Spatial/Temporal Subband Decomposition

An adaptive wideband beamforming structure with a lower computational com- plexity and higher convergence speed is proposed, which is achieved by combined spatial/temporal subband decomposition. First, the received array signals are processed by a transformation ma- trix, and then split into subbands by a series of analysis fllter banks. At each subband, an independent beamformer can be operated, where some of the lower subbands can be discarded without afiecting its performance due to the highpass flltering efiect of the transformation matrix. 1. INTRODUCTION Beamforming has found many applications in various areas ranging from sonar and radar to wireless communications (1). It is a signal processing technique to form beams in order to receive signals illuminating a sensor array from speciflc directions, whilst attenuating signals from other directions. In a statistically optimum beamformer (2), since the statistics of the array data are often not known or may change over time, adaptive algorithms may be used to determine the beamformer's coe-cients. A beamformer structure for wideband array signals is shown in Fig. 1, where each of the M received signals is processed by an FIR fllter with a length of J and the outputs of these FIR fllters are then sumed up to form the flnal output. To perform adaptive wideband beamforming with high interference rejection and angular resolu- tion, arrays with a large number of sensors and fllter coe-cients have to be employed. Reducing the resultant high computational complexity and increasing its convergence speed has motivated many difierent solutions, including partially adaptive beamforming (3), transform-domain or frequency- domain beamforming (4,5), and subband beamforming (6,7). Recently, a generalised sidelobe can- celler (GSC) with combined subband decomposition in both the temporal and spatial domains was proposed in (8), where the columns of the blocking matrix are judiciously designed to generate a highpass flltering efiect to the array signals. The bandlimited spectra of the resultant outputs are then exploited by subband decomposition and appropriately discarding the low-pass subbands, where there is no signal existing. In this paper, we generalize the idea in (8) and propose a subband-selective transformation matrix, which is applied to the received array signals as a preprocessing step. Each of the outputs