Using the speech intelligibility index for evaluating speech test results and hearing aid characteristics

Abstract

Understanding speech is crucial for human communication. Therefore, speech audiometry plays an important part in the assessment of the hearing impaired, and improved speech intelligibility is a major goal of hearing aid fitting. The aims of this study were to improve speech test procedures and to validate the speech Intelligibility Index (SII) as a means for predicting speech test results and evaluating hearing aid characteristics.The established Swedish Phonemically Balanced (SPB) word lists were mixed with speech-weighted noise, and normative performance data were obtained with normal-hearing subjects. To facilitate the assessment of individual speech recognition results, functions were derived that related the SII to measured speech recognition scores. A prediction procedure, in which the standard SII was complemented with corrections for supra-threshold deficits, was evaluated on a group of hearing-impaired persons. The modified SII was also evaluated as a means for comparing hearing aid characteristics. Predicted speech recognition based on the SII was compared with speech recognition scores, just-follow-conversation results and subjective judgments of speech intelligibility for different conditions of amplification and background noise.The speech-to-noise ratio +4 dB was found to be appropriate for clinical use. A pre-mixed test material, called SPBN (Swedish Phonemically Balanced words in Noise), was recorded on CD. Speech recognition scores of hearing impaired persons were on average well predicted with the modified SII, though there were considerable differences among individuals, which confirmed the need for speech recognition tests. A computer program was developed to calculate predicted performance on the SPBN test. When the modified SII was applied for evaluating hearing aids, the calculated intelligibility showed an overall good agreement with measured speech recognition results and also with subjectively judged speech intelligibility. The within subject standard deviation of the predicted scores was estimated to be about 1%. Conclusions: The SPBN test is suitable for every-day clinical determination of speech recognition scores. The SII based prediction procedure facilitates evaluation of individual speech recognition results. The modified SII can be used in hearing aid fitting procedures in order to evaluate frequency-gain characteristics of hearing aids in terms of expected speech intelligibility in background noise.