On the Use of Auditory Spectral Representations for the Comparison of Hearing Compensation Techniques

An objective method to give a reliable prediction of which compensation strategy should behave better for a given hearing loss is proposed. This method is based on the estimation of the Auditory Spectral Representations (ASRs) and the goal is to assess which compensation strategy produces ASR very similar to that generated in a normal hearing subject. ASRs are the representations of a speech waveform in the cochlea, after auditory periphery processing. The assumption underlying the current analysis is that the closer the 'distance' between ASRs obtained in a normal and in an impaired listener after compensation, the more likely to achieve a better speech perception in the impaired listener. The 'distance' between the ASR of a normal and an impaired listener is evaluated for sloping sensorineural hearing loss of different degrees (mild, moderate, and severe) and two compensation strategies (NAL-R and FIG6). As a general remark, it was clear that the distance between the ASR of the normal the impaired condition increased with hearing loss for both strategies. This would suggest that, as the degree of hearing loss increases, the compensated speech perception progressively deteriorates for both strategies. More interestingly, simulations showed that, for a given hearing loss degree, the two strategies produced different results: FIG6 generated ASR much more similar to the reference than NAL-R for all hearing loss profiles but the severe one. For this latter profile, NAL-R, generated ASR much more similar to the reference than FIG6. This seems to indicate that the use of a nonlinear strategy, such as FIG6, in severe hearing loss might generate more distortions than NAL R.