Noise reduction for digital hearing aids

Abstract

The normal-hearing system extracts monaural and binaural features from the signals at the left and right ears in order to separate and classify sound sources. Robustness of source extraction is achieved by exploiting redundancies in the source signals (auditory scene analysis). ASA is closely related to the ‘‘Cocktail Party Effect,’’ i.e., the ability of normal-hearing listeners to perceive speech in adverse conditions at low signal-to-noise ratios. Hearing-impaired people show a reduced ability to understand speech in noisy environments, stressing the necessity to incorporate noise reduction schemes into hearing aids. Several algorithms for monaural, binaural and multichannel noise reduction have been proposed, which aim at increasing speech intelligibility in adverse conditions. A summary of recent algorithms including directional microphones, beamformers, monaural noise reduction and perceptual model-based binaural schemes will be given. In practice, these schemes were shown to be much less efficient than the normal-hearing system in acoustically complex environments characterized by diffuse noise and reverberation. One reason might be that redundancies in the source signals exploited by the hearing system are not used so far by noise reduction algorithms. Novel multidimensional statistical filtering algorithms are introduced that might fill this gap in the future. [Work supported by BMBF 01EZ0212.]