Determining perceived sound quality in a simulated hearing aid using the international speech test signal

Based on speech segments from female talkers of different languages, the International Speech Test Signal (ISTS) has speech-like acoustic properties but is not intelligible. This study investigated whether sound quality ratings for the ISTS are similar to ratings obtained for linguistically meaningful speech.

A simulated hearing aid was used to process the ISTS for a variety of noise, nonlinear, linear, and combined nonlinear and linear processing conditions. The sound quality of the test conditions was rated by both normal-hearing and hearing-impaired (HI) listeners. Listeners' ratings were then modeled using the Hearing Aid Sound Quality Index (HASQI).

The HASQI predictions indicated how closely the ISTS sound quality ratings were to listener ratings of American English. In the normal-hearing group, correlations between HASQI predictions and listener ratings were 0.90 for nonlinear processing, 0.78 for linear processing, and 0.88 for combined processing. In the HI group, correlations between HASQI predictions and listener ratings were 0.96 for nonlinear processing, 0.94 for linear processing, and 0.96 for combined processing.

Since the ISTS is intended for use with hearing aids, the accuracy of the HASQI predictions for the HI listeners reinforces the validity of using this signal for hearing aid sound quality ratings and predictions.