Reduced-bandwidth and low-complexity multichannel wiener filter for binaural hearing aids

Binaural hearing aids use a wireless link to exchange the signals received at each side, allowing the implementation of sophisticated noise-reduction strategies. Although different noise-reduction strategies have been reported for binaural hearing aids, those based on Multichannel Wiener filter (MWF) have been identified to provide the best SNR improvement and output sound quality. The real-time implementation of these MWF techniques involves different challenges such as reduction of computational complexity, processing delay, and transmission bandwidth, among others. The MWF methods have been generally derived in the FFT domain. This paper explores a different approach, where the FFT is replaced by an auditory representation. For this representation, two strategies are analyzed: IIR filter-banks and wavelet packets (WP). The WP representation is shown to provide reduction of the computational complexity, processing delay, and transmission bandwidth, without degrading the performance of SNR improvement and sound quality.