Abstract:

The amount of acoustic information available from the speech signal was reduced by removing all regions of the waveform in which the rms amplitude fell below a threshold value. This removal was accomplished by envelope center clipping followed by attenuation and envelope expansion. The processing was carried out independently in three-frequency bands, which were then recombined. The result was intended to simulate the effects of hearing loss with recruitment. Phoneme recognition in consonant--vowel--consonant words was measured in eight normally hearing subjects as a function of threshold. Subjects, essentially, heard the top x dB of the speech signal, where x varied from 3--30 dB. Significant individual differences of performance were found, covering a range of (plus or minus)3.6 dB when measured in terms of the threshold required for a score of 50%. Because all subjects in this experiment had access to the same sensory evidence, the findings support a perceptual closure explanation of individual differences. It may be, therefore, that some of the individual differences of phoneme-level speech perception by the hearing impaired, or by normals listening in noise, also reflect differences in the ability to complete perceptual closure. [Research supported by NIDCD.]