Predictions of speech quality under conditions of noise and distortion

Abstract

Noise and distortion reduce speech intelligibility and quality in audio devices such as hearing aids. The purpose of this study is to quantify and model the effects of signal degradation on the perception of speech quality in normal-hearing and hearing-impaired listeners. The stimuli were sentences subjected to eight different levels of additive noise, peak clipping, and center clipping distortion. The subjects listened to all possible comparisons of pairs of the 24 degraded sentences, and for each comparison indicated their preference. Multi-dimensional analysis is used to model the subject decision spaces, and the data are also modeled using a one-dimensional metric to predict the subject quality judgments. The one-dimensional metric is an extension of the procedure developed by Kates and Arehart [J. Acoust. Soc. Am. 117, 2224–2237 (2005)] to model speech intelligibility. That approach is based on dividing the speech signal into three amplitude regions, computing the coherence for each region, and then combining the three coherence values across frequency in a calculation based on the speech intelligibility index (SII). The one-dimensional and multi-dimensional analyses are able to accurately model the quality perception of both normal-hearing and hearing-impaired listeners. [Work supported by GN ReSound and Centers for Disease Control.]