Predicting binaural effects in perceived listening effort
* Presenting author
Recent studies measured perceived listening effort in addition to speech intelligibility. The main interest in listening effort is because it allows assessing differences between listening conditions at which intelligibility is at ceiling. In this light, it would be desirable to have quantitative prediction models to instrumentally compute listening effort. However, previous studies investigating predictors of perceived listening effort have focused on monaural conditions, thus neglecting the important advantage humans can obtain when target and maskers are spatially separated. This contribution therefore exploits a recent data set of binaural listening effort as a basis for extending and testing a binaural prediction model. The front-end of the model is the equalization-cancelation stage to model binaural processing. The back-end is the speech intelligibility index (SII). The presented model extension consists of deriving a mapping function relating SII and perceived listening effort for a diotic, anechoic reference condition, and then using the same mapping function to predict all other conditions. The results show that this relatively simple approach can predict the experimental data with good accuracy. In particular, it generalizes well to reverberant conditions as well as binaural unmasking conditions, indicating that this model approach may be a useful tool for assessing speech perception.