Contribution
Speech intelligibility as a measure for vehicle interior sound quality
* Presenting author
Abstract:
Different measures such as weighted sound pressure levels or psychoacoustic loudness are widely used for evaluating the interior sound quality of a vehicle during the development process. Using speech intelligibility predictions could possibly provide a more customer-oriented evaluation method, especially for driving conditions above approximately 80km/h. A common, standardized method for these predictions is provided by the Speech Intelligibility Index (SII, ANSI/ASA S3.5). The calculated index already gives a measure to compare different vehicles but does not provide information for the development process. However, since the calculation algorithm includes a filterbank, frequency bands in the vehicle sound that show a high speech-masking potential could be identified. To allow for a more in-depth analysis of a given audio signal, a similar method was developed using a speech corpus containing different speakers applying several levels of vocal effort to their spoken sentences. On this basis, a masking analysis can be carried out along time segments and across frequency bands allowing for slightly more detailed observations during the development process. These methods are compared against each other and are discussed in comparison with more direct approaches such as weighted sound pressure levels or loudness.