A Joint Auditory Attention Decoding and Adaptive Binaural Beamforming Algorithm for Hearing Devices

Traditional adaptive binaural beamforming algorithms for hearing devices often assume that the target talker is known or can be derived from the listener’s look direction. When this assumption is violated, the traditional beamforming algorithms often produce distorted target speech and less than optimal noise and interference suppression. Recent advances in electroencephalography (EEG) and its applications to auditory attention decoding have offered a potential solution for tracking the listeners auditory attention in a multi-talker environment [2]–[5]. In this paper, we propose a unified model for joint auditory attention decoding and adaptive binaural beamforming, and solve the problem using an iterative optimization approach. The proposed algorithm has two advantages over the existing algorithms. First, the optimization objective aims to balance auditory attention alignment, target speech distortion, noise and interference suppression. Secondly, there is no need to estimate the speech envelope of each talker from the noisy and reverberant mixture which is a very challenging problem in practice. The proposed algorithm was evaluated using a newly recorded EEG database for a multi-talker, noisy and reverberant environment [6]. The evaluation results confirm the benefits of the proposed algorithm.