Abstracts of the International Hearing Aid Research Conference 2008 pages:78-79
International Hearing Aid Research Conference 2008 location:California, USA date:16-20 August 2008
Multi-microphone noise reduction algorithms are commonly implemented in modern hearing aids to improve speech intelligibility in noisy environments. The development of these algorithms has mostly focused on monaural systems. The human auditory system is a binaural system which compares and combines the signals received by both ears to perceive and localize a single sound source. Providing two monaural, independently operating, noise reduction systems (a bilateral configuration) to the hearing aid user may disrupt binaural information, needed to localize sound sources correctly.
In this study, three multi-microphone noise reduction algorithms were evaluated with respect to their influence on speech intelligibility and on the ability to localize sound sources. This was done using theoretical, objective and perceptual evaluations in different spatial scenarios. Two recently developed noise reduction techniques for binaural hearing aids were evaluated, namely the binaural multichannel Wiener filter (MWF) and the binaural multichannel Wiener filter with partial noise estimate (MWF-N). The binaural MWF theoretically preserves the binaural cues of the speech component. To preserve the binaural cues of both the speech and the noise components, the MWF-N was developed. This algorithm, in theory, sacrifices some noise reduction to preserve the binaural cues of the noise component. During the different evaluations, a bilateral adaptive directional microphone (ADM) was used as a reference system since it is widely used in commercial hearing aids.
The main conclusions are: a) The ADM only preserves localization in the forward direction. In these directions limited or no speech-in-noise enhancement is obtained. b) The MWF preserves localization of the target speech component but can distort localization of the noise component. Objective and perceptual evaluations showed that these distortions are often smaller than those predicted by theory. Moreover, they are dependent on signal-to-noise ratio and masking effects. By adding more contralateral microphone signals to the binaural MWF, the noise reduction performance significantly improved c) The MWF-N improved the ability to localize the noise component when compared with the MWF. Objective performance measures showed that this came at the cost of noise reduction. However, perceptual evaluations did not show this tendency. When speech and noise components are spatially well separated, the MWF-N even outperformed the MWF in terms of speech intelligibility. This can be explained by an increased release from masking when preserving the binaural cues of both the speech and noise components.