The Hearing-Aid Speech Quality Index (HASQI)

Signal modifications in audio devices such as hearing aids include both nonlinear and
linear processing. An index is developed for predicting the effects of noise, nonlinear
distortion, and linear filtering on speech quality. The index is designed for both normalhearing
and hearing-impaired listeners. It starts with a representation of the auditory
periphery that incorporates aspects of impaired hearing. The cochlear model is followed
by the extraction of signal features related to the quality judgments. One set of features
measures the effects of noise and nonlinear distortion on speech quality, whereas second
set of features measures the effects of linear filtering. The hearing-aid speech quality
index (HASQI) is the product of the subindices computed for each of the two sets of
features. The models are evaluated by comparing the model predictions with quality
judgments made by normal-hearing and hearing-impaired listeners for speech stimuli
containing noise, nonlinear distortion, linear processing, and combinations of these signal
degradations.