Design of two syllabic nonlinear multichannel signal processors and the results of speech tests in noise.

OBJECTIVE

Multichannel syllabic compressors have not yet shown clear advantages for speech perception. New multichannel syllabic nonlinear processors are designed and evaluated in this study to test whether they enhance speech perception scores.

DESIGN

Nonsense consonant-vowel-consonant words have been processed in real time with two syllabic nonlinear methods in nine different frequency channels: 1) 30 dB of speech information is mapped into the residual dynamic range (DR) of the subjects with hearing impairment using either compression or expansion; 2) speech levels above the rms level are compressed with a ratio of 3:1, and input levels below the rms level are mapped into the reduced DR of the subjects. The overall frequency spectrum of speech is tilted with more high-frequency output for this condition. The two syllabic nonlinear processors were compared with two linear reference systems (one of them with the same spectral tilt as the second nonlinear condition) for seven hearing-impaired subjects.

RESULTS

On average, the syllabic nonlinear processors show scores similar to one of the linear systems. The linear reference with tilt has, on average, 3% higher scores, mainly due to improved vowel identification.

CONCLUSIONS

No negative effects were shown on average for the syllabic nonlinear processors, but no positive effects were demonstrated either. Tilting of the overall speech spectrum was advantageous.