Speech understanding in noise: Contributions of compressive nonlinearities in the basilar-membrane response

Abstract

The contribution of compressive nonlinearities in the basilar-membrane response was assessed by measuring speech recognition in noise as a function of noise level and growth of masking for tones. Consonant recognition was measured in interrupted noise at overall masker levels of 47–77 dB SPL. Additionally, thresholds for a 10-ms, 2.0-kHz tone were measured in a masker ranging in level from 40 to 85 dB SPL and centered at 1 kHz. Subjects were younger and older adults with normal hearing. With speech at a constant level and the masker level varying, it was hypothesized that the response of the basilar membrane to the masker would be linear at lower levels and compressed at medium to higher levels, resulting in less effective masking at higher masker levels. Consistent with the hypothesis, differences between observed and predicted consonant-recognition scores, determined using AI calculations, varied as a function of masker level, especially for younger subjects. At lower masker levels, observed scores declined as predicted as masker level increased; with further increases in masker level, scores declined less than predicted. For tonal growth-of-masking functions, breakpoints were correlated with quiet thresholds: both were significantly higher for older than younger subjects. [Work supported by NIH/NIDCD.]