Change Prediction For Low Complexity Combined Beamforming And Acoustic Echo Cancellation

Time-variant beamforming (BF) and acoustic echo cancellation (AEC) are two techniques that are frequently employed for improving the quality of hands-free speech communication. However, the combined application of both is quite challenging as it either introduces high computational complexity or insufficient tracking. We propose a new method to improve the performance of the low-complexity beamformer first (BF-first) structure, which we call change prediction (ChaP). ChaP gathers information on several BF changes to predict the effective impulse response seen by the AEC after the next BF change. To account for uncertain data and convergence states in the predictions, reliability measures are introduced to improve ChaP in realistic scenarios.