Speech recognition performance and predictions: bridging acoustics and audiology
* Presenting author
The auditory system has a remarkable ability to perform well in acoustically challenging conditions. This ability worsens with increasing age and hearing deficits. Studies related to better understanding and improvement of speech recognition in diverse acoustic conditions for listeners with different hearing status will be a leitmotiv of this talk. First, the matrix sentence test as a reliable, internationally compatible tool for speech recognition assessment will be introduced. Examples of application of the matrix type tests in basic research studies as well as in hearing diagnostics will be presented. Furthermore, the outcomes of the listening experiments will be compared to the predictions of “effective” speech intelligibility models like a binaural speech intelligibility index-based model (BSIM) or an automatic speech recognition-based model (FADE). They enable assessing speech recognition in a quick, inexpensive, and reproducible manner. Finally, applications of this kind of models for assessing the acoustic treatment of rooms or speech enhancement techniques in terms of speech recognition will be shown. These examples demonstrate a bridge between individual hearing capabilities (measured with audiological test methods) and room acoustics with the aim to improve the acoustic environments particularly for vulnerable persons, i.e. school children and handicapped persons.