Influence of simulated spatial conditions and reverberation on speech intelligibility prediction accuracy with the Simulation Framework for Auditory Discrimination Experiments (FADE)

Considering complex acoustic scenes in rehabilitative audiology and hearing device assessments requires understanding the influence of multiple interacting factors. Speech intelligibility models provide a systematic way to explore and predict these effects. However, they must be able to deal with acoustic conditions including different numbers and spatial configurations of sound sources, and the presence of reverberation. Here, the binaural implementation of the automatic speech recognition-based approach Simulation Framework for Auditory Discrimination Experiments (FADE) was validated for predicting speech intelligibility in ten acoustic scenes of varying spatial complexity without and with reverberation. The outcomes were compared with data from normal-hearing listeners and with the predictions of an index-based, intrusive model, the Binaural Speech Intelligibility Model (BSIM). The interferer constellation in anechoic conditions had a large effect on speech recognition thresholds, which was predicted with a high accuracy by both models (0.90 ≤ R² ≤ 0.97 between modeled and predicted data). This spatial influence was greatly reduced in reverberant conditions, which was correctly predicted by the models. Reverberation had a detrimental effect on speech intelligibility, with larger changes in measured than in predicted data. In conclusion, the study demonstrates that FADE is able to predict the influence of interferer constellation for normal-hearing listeners even in complex conditions and performs comparably to the well-established BSIM, without the need to separate speech and noise inputs or a reference. However, FADE needs improvement to correctly predict the detrimental effect of reverberation on speech intelligibility.