Values of the speech intelligibility index (SII) were found to be different for the same speech intelligibility performance measured in an acoustic perception jury test with 35 human subjects and different background noise spectra. Using a novel method for in-vehicle speech intelligibility evaluation, the human subjects were tested using the hearing-in-noise-test (HINT) in a simulated driving environment. A variety of driving and listening conditions were used to obtain 50% speech intelligibility score at the sentence Speech Reception Threshold (sSRT). In previous studies, the band importance function for ‘average speech’ was used for SII calculations since the band importance function for the HINT is unavailable in the SII ANSI S3.5-1997 standard. In this study, the HINT jury test measurements from a variety of background noise spectra and listening configurations of talker and listener are used in an effort to obtain a band importance function for the HINT, to potentially correlate the calculated SII scores with the measured speech intelligibility scores. If a mathematical solution for such a problem is not found, then the measured speech intelligibility at the sSRT, considering different background noise spectra, is not, as expected, associated with a particular SII score; for example, 0.5. As a result, SII at sSRT may not be a suitable method for future in-vehicle speech intelligibility evaluation or target-setting, as it is sensitive to the changes in (speech) signal to (background) noise ratio and/or background noise spectrum shape.
