Predicting effects of impaired cochlear processing on consonant discrimination in stationary noise

Cochlear hearing loss is typically associated with reduced sensitivity due to inner hair-cell (IHC) and outer hair-cell (OHC) dysfunction. OHC dysfunction also leads to supra-threshold deficits, such as reduced basilar-membrane (BM) compression as well as reduced frequency selectivity and temporal resolution.

Listeners with a cochlear damage typically have difficulties with speech understanding in the presence of background noise. In this study, the goal was to investigate the relation between individual consonant confusions in stationary noise and deficits in cochlear signal-processing as characterized by the audiogram and estimates of the BM input-output characteristics.

Cochlear processing in individual listeners was simulated using a computational model of auditory signal processing and perception (CASP) and was used as a front end in a consonant discrimination system. Individual error patterns from a Diagnostic Rhyme Test (DRT) were measured and analyzed in terms of acoustic-phonetic features.

This was done for three listeners with cochlear hearing loss and at two signal-to-noise ratios. It is shown that the predicted errors patterns matched the measured patterns in most conditions. Thus, an incomplete representation of the speech sounds due to deficits in cochlear processing could be related to the performance in the speech perception task.

In addition, it was studied to what extent the data could be accounted for based on reduced sensitivity only – assuming that BM compression, frequency selectivity and temporal resolution are the same as in normal-hearing listeners. For two out of the three listeners, the supra-threshold deficits needed to be included in order to account for the data, while for the third listener, the predicted error rates were similar for the two model versions.

Overall, the results suggest a clear relation between deficits in cochlear signal processing and consonant identification error patterns, and indicate that the supra-threshold deficits associated with a cochlear damage need to be taken into account. The findings might be interesting for applications, such as the evaluation of hearing-instrument signal processing, where the effects of specific processing strategies can be simulated for individual hearing losses.