Blind Prediction of Speech Recognition Thresholds in Binaural Listening Conditions for Individual Hearing-Impaired Listeners
The prediction of speech recognition thresholds (SRT) for hearing-impaired listeners in realistic acoustic environments (simulated or real-world conditions) is a highly relevant topic. The contralateral inhibition (KAIN) approach within the Framework for Auditory Discrimination Experiments (FADE) has recently been shown to provide promising predictions with normal-hearing listeners for binaural masking level differences and SRTs in realistic listening conditions. A special property of the approach, in contrast to other models of binaural speech perception, is that it works blindly, i.e., FADE KAIN is able to educe relevant information from the speech by the difference of extracted Gabor features from the binaural channels distorted by a supra-threshold-noise component. Thus, the model is not dependent on information about the direction of the signal and masker sources. Even though FADE is able to model hearing loss as well, the prediction accuracy of the FADE KAIN model has not been further evaluated yet for hearing-impaired listeners. In this contribution, binaural SRT predictions using FADE KANE for individual hearing-impaired listeners will be presented. The influence of different supra-threshold noise components within the model will be considered and evaluated. Predictions of the model will be compared to empirical data from the literature and will be discussed further.