Combined Noise Reduction and Coherence Reshaping for Binaural Hearing Aids

Noise reduction algorithms in hearing aids are crucial to improve speech understanding in background noise for hearing impaired persons. For binaural hearing aids, algorithms that exploit the microphone signals from both the left and the right hearing aid are considered to be promising techniques for noise reduction, because in addition to spectral information spatial information can be exploited [1]. In addition to reducing noise and limiting speech distortion, another important objective of binaural noise reduction algorithms is the preservation of the listener’s impression of the acoustical scene, in order to exploit the binaural hearing advantage and to avoid confusions due to a mismatch between the acoustical and the visual information. This can be achieved by preserving the binaural cues of the speech and the noise component. To achieve binaural cue preservation, two main concepts for binaural noise reduction have been developed. In the first concept, the multi-channel signals are used to calculate a real-valued gain, where the same gain is applied to the reference microphone in the left, respectively right hearing aid [2]. This processing strategy allows perfect preservation of the binaural cues of both the speech and the noise component, but typically suffers from limited noise reduction performance and possible single-channel noise reduction artifacts. The second concept is to apply a complex-valued filter to all available microphone signals on the left and the right hearing aid, combining spatial and spectral filtering. Using this processing strategy, a large noise reduction performance can be achieved, but the binaural cues of the residual noise component are not guaranteed to be preserved. In [1] the binaural Speech Distortion Weighted Multi-channel Wiener Filter (MWF) has been presented. It has been theoretically proven in [3] that in case of a single speech source this technique preserves the binaural cues of the speech component but typically distorts the binaural cues of the noise component. Hence, algorithms have been proposed that aim to preserve the binaural cues of directional noise sources by adding a cue preservation term related to the Interaural Transfer Function (ITF), the Interaural Level Difference (ILD) or the Interaural Time Difference (ITD) to the basic noise reduction cost function [3, 4, 5]. In contrast to directional noise sources, the spatial characteristics of e.g. spatially isotropic noise however can not be properly described by the ITF, but rather by the Interaural Coherence (IC). In this paper we propose an extension of the MWF with a term related to the IC preservation of the noise component. Experimental results for two scenarios show that the proposed algorithm yields a good preservation of the Interaural Coherence without significantly degrading the output SNR compared to the binaural MWF.