Multichannel syllabic compression for severely impaired listeners

Two listeners with congenital hearing losses characterized by flat audiograms and dynamic ranges of 18-33 dB were tested with three compression systems and one (reference) linear amplification system. The compression systems placed progressively larger amounts of speech energy within the listener's residual dynamic range, by raising to audibility and compressing 25, 50, and 90 percent of the short-term input amplitude distribution in each of 16 frequency bands. The comparison linear system was defined by adjusting six octave-wide bands of speech to comfortable levels. System performance was evaluated with nonsence CVC syllables presented at a constant input level and spoken by two talkers. Extensive training was provided to ensure stable performance. The results were notably speaker-dependent, with compression consistently providing better performance for one speaker, linear amplification for the other. Averaged over speakers, however, there was no net advantage for any of the compression systems for any listener. The use of high compression ratios and large input ranges tended to degrade perception of initial consonants and vowels. Under some conditions, however, final consonant scores were higher with compression than with linear amplification. Compression generally enhanced the distinction between stops and fricatives, but degraded spectral-concentration and relative-intensity cues required to identify place of articulation.