Research On Speech Intelligibility In Different Sound Fields And Its EEG

The presence of early reflections and late reflections has impacts on SI,which depends on several factors.Speech and noise in separate positions and reverberation time can affect the effect of reflections on speech intelligibility.In this work,we measured speech reception threshold to assess the impact of three types of sound field conditions(direct sound,direct sound and early reflections,and binaural room impulse response)on SI with different spatial separated speech and noise and reverberation times.Besides,Both the inter-aural crosscorrelation coefficient and the better ear speech transmission index were calculated to reflect the changes of SI.In addition,the event-related potentials evoked by three types of sound field conditions at different reverberation times were recorded by EEG experiments.First,speech reception thresholds were measured in three types of sound fields with spatially separated speech and noise.The results show that early reflections reduce SI compared to the case of only direct sound when the target speech signal and noise masking signal are always in the same sound field.The improvement of SI with direct sound only is much larger than that with direct sound and early reflections and binaural room impulse response.Second,speech reception thresholds of three types of sound field conditions with different reverberation times were measured,and the brain wave changes induced by the stimulus signals under three types of sound field conditions with different reverberation times were recorded by the EEG experiment.The results show that:(1)The amplitude peak and latency of P300 induced by a combination of the direct sound and early reflections,as well as the complete binaural impulse response is significantly larger than that induced by the direct sound only.There is a tendency that the stimuli signal with longer reverberation time and more reflection components induces a larger amplitude peak of P300;(2)There is a tendency that the stimuli signal with longer reverberation time and more reflection components evokes larger amplitude peak and longer latency of mismatch negative;(3)The amplitude peak of P300,the peak amplitude and latency of mismatch negative are all closely correlated with the measurement results of the speech reception threshold.The research in this paper not only helps to understand the influence of early,late reflections and noise on speech intelligibility but also further reflects the brain response mechanism in reflection perception and provides references for the design and optimization of speech intelligibility prediction algorithms.