Robust Adaptive Wideband Beamforming Against Direction and Sensor Location Errors

The performance of the existing robust beamformers can be still degraded by the bias between the nominal steering vector and the actual one. In this paper, a novel robust wideband beamformer based on the time-frequency distributions is proposed, which can estimate the steering vector accurately even in the presence of direction and sensor location errors. Firstly, it develops an approach for wideband signals to select the single-source auto-source time-frequency (TF) points of the source signals. Then these TF points are utilized to obtain the steering vectors without using the perturbed array manifold and direction information. Finally, a higher output signal-to-interference-plus-noise ratio (SINR) is achieved for the minor bias between the estimated steering vectors and the actual ones. Simulation results demonstrate that the proposed algorithm outperforms other conventional robust beamforming approaches and can achieve high output SINR close to the ideal beamformer over a broad range of direction and sensor location errors.