A fundamental property of the auditory system is that it operates as a kind of frequency analyzer. This property has been mainly associated with the ear's ability to analyze spectral shapes. Consistent (i) with the finding that the ear's sensitivity in frequency is much finer than would be required to understand speech in quiet, and (ii) with a recent view of speech recognition in noise suggesting that speech sounds are identified based on samples from a limited number of """"""""time-frequency regions which contain a reasonably undistorted view of local signal properties"""""""", it is proposed that the primary function of the ear's fine spectral resolution is to segregate speech from noise. The present project will investigate the role of the normal peripheral auditory system in the unmasking of speech. One of the primary aims is to determine the amount of information needed to reconstruct an interpretable representation of the target speech. Other major aims include determining the amount of information potentially available at the output of each auditory filter and examining how listeners determine which auditory filters convey relevant speech information in the presence of background noise. The method used to achieve the aims proposed here will primarily involve measuring speech recognition in noise in normal-hearing listeners while manipulating the available number of auditory filter outputs and the signal-to- noise ratio within each band. The long-term goal of the present project is to clarify the mechanisms underlying speech perception in noise by the normal auditory system. The novel view proposed here, however, has the potential to further our understanding of the deficits in impaired listeners. Listeners with hearing loss of cochlear origin often have considerable difficulty understanding speech in noise. Reduced frequency selectivity is one of the deficits frequently evoked to account for this difficulty. It is assumed that broader auditory filters smooth the representation of the speech spectrum and that background noise exacerbates this smoothing by reducing the prominence of spectral peaks. While NH listeners may decompose the incoming signal into as many as 30 bands in the range 80 Hz to 8000 Hz, listeners with moderate to severe hearing loss presumably rely on as few as 10 bands if a broadening factor of 3 is assumed. Moreover, broader filters pass more noise, further diminishing the probability for any given band to convey undistorted speech. Accordingly, reduced frequency selectivity can be viewed as diminishing the probability to uncover regions in which the target signal is least affected by the background. The failure of current rehabilitation devices to fully restore speech intelligibility in noise may arise from their purely analytical approach of the role of the peripheral auditory system. A new approach, taking into consideration the noise reduction capability of the cochlea, may have considerable practical implications for the design of improved cochlear implants and hearing aids. Public Health Relevance: Nearly 35 million Americans suffer from measurable hearing impairment and related speech disorders and 2 million are completely deaf. Current rehabilitation devices fail to fully restore speech intelligibility in these listeners, especially when background noise is present. The novel view of the role of the peripheral auditory system proposed in this project will lead to a better understanding of speech processing deficits in impaired listeners, and therefore has the potential to provide considerable practical implications for the design of improved cochlear implants and hearing aids.
Nearly 35 million Americans suffer from measurable hearing impairment and related speech disorders and 2 million are completely deaf. Current rehabilitation devices fail to fully restore speech intelligibility in these listeners, especially when background noise is present. The novel view of the role of the peripheral auditory system proposed in this project will lead to a better understanding of speech processing deficits in impaired listeners, and therefore has the potential to provide considerable practical implications for the design of improved cochlear implants and hearing aids.
