Extending a computational model of auditory processing towards speech intelligibility prediction

Activity: Talks and presentationsConference presentations


A speech intelligibility model is presented, based on the computational auditory signal processing and perception model (CASP; Jepsen et al., 2008). CASP has previously been shown to successfully predict psychoacoustic data of normal hearing (NH) listeners obtained in conditions of, e.g., spectral masking, amplitude-modulation detection, and forward masking (Jepsen et al., 2008). Furthermore, CASP can be tuned to model data from individual hearing-impaired listeners in different behavioral experiments (Jepsen and Dau, 2011). In this study, the CASP model is investigated as a predictor of intelligibility for Danish sentences for NH listeners.
The model receives the clean and degraded speech as input. The signals are processed through outer- and middle-ear filtering, a non-linear auditory filterbank (DRNL, López-Poveda and Meddis, 2001), adaptation loops, and a modulation filterbank. The internal representations produced at the end of these stages are analyzed using a correlation-based back end.
Here, predictions of speech intelligibility obtained with the speech-based CASP implementation are presented and compared to speech intelligibility data measured in conditions of additive noise, phase jitter, spectral subtraction, ideal binary mask processing and reverberation.
Event titleInternational Symposium on Auditory and Audiological Research
Event typeConference
LocationNyborg, Denmark