Abstract
Speech from a target speaker reaches a listener via multiple paths in a room due to room reflections. As sound waves from the target speaker approach the listener, degradation to the signal is caused by ambient noise and reverberant energy. The speech transmission index (STI) is a commonly used metric for predicting speech intelligibility accounting for both noise and reverberation. This metric, however, is a monophonic measure that does not take into consideration binaural cues used for unmasking undesired effects. As a result, using the STI on its own tends to under-predict intelligibility under binaural listening conditions. The proposed research aims to improve speech intelligibility predictions with the presence of room effects by implementing a psychophysical binaural model as a front-end to the STI calculation. The equalization-cancellation (EC) theory is applied to spatially unmask noise, while late incoherent reverberant energy is suppressed by applying a weighting function based on interaural coherence. Preliminary comparisons between listening tests and model predictions reveal promising results, indicating a useful tool in acoustical planning in addition to further study into binaural suppression processes.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call