A Low-Complexity Adaptive Beamformer for Joint Reverberation and Noise Suppression

Reverberation and background noise can severely affect the quality and intelligibility of recorded speech, potentially impairing speech communication and human-machine interaction systems. The minimum variance distortionless response (MVDR) beamformer is commonly used to jointly suppress reverberation and noise, but it is computationally intensive due to the need for matrix inversion at each time frame and subband. In this paper, we introduce a beamforming method that combines an MVDR beamformer optimized for noise reduction with a fixed maximum directivity-factor beamformer for reverberation suppression. This hybrid beamformer offers a more efficient implementation than the traditional MVDR beamformer and adapts to varying levels of reverberation and noise by adjusting a single weighting factor. Simulation results demonstrate that the proposed beamformer achieves significantly faster processing speeds compared to the optimal MVDR beamformer, with only minor performance degradation in noise reduction and reverberation suppression.