Noise reduction for "intelligent" hearing aids

One major problem of people with hearing impairment is the 'Coctail-Party-Effect', i.e., the reduced ability to understand speech in noisy environments. Normal hearing listeners use binaural information, i.e., interaural level and phase/time differences in addition to monaural cues to localize sound sources in the environment. Based on the localization result and on the monaural cues, the listener performs a significant noise reduction by concentrating on one of the sources. Noise reduction algorithms have been proposed that try to imitate some aspects of the (binaural) information processing in the human hearing system by using two microphones at both ears and a central processor. Based on a two channel input, it is theoretically possible to cancel one interfering directional noise source at a time. Due to the relatively large distance between microphones, it should additionally be possible to detect and cancel reverberation based on coherence measures. In practice however, 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 the hearing system additionally exploits redundancies in the source signals to increase the robustness of source separation (‘Auditory Scene Analysis’), which are not used so far by noise reduction algorithms. Novel multidimensional statistical filtering algorithms are introduced that might fill this gap in the future.