Further investigations on binaural intelligibility of reverberant speech

Abstract

There is a binaural intelligibility advantage for many subjects when listening to reverberant speech. Two processes offer explanations as to why the binaural advantage exists. First, directional information may identify echoes and remove a portion of the reverberation in a manner similar to the precedence effect. A more likely possibility is that the reverberation behaves as additive noise with a specific time and frequency dependence. In this case, subjects may utilize the coherence of the left and right channels to limit the effects of the incoherent reverberation, similar to binaural masking release. To test the latter hypothesis, subjects listen to diotic (L+R), binaural (L&R), and diotic (L or R) reverberant words. The diotic treatment (L+R) adds two recordings measured at the location of the listener’s ears. Since the listener faces the source, the direct signals are coherent while some cancellation is expected for the reverberation. The presentation will compare diotic (L+R) intelligibility to the binaural (L&R) condition as well as the diotic (L or R) control condition where the left or right signal plays to both ears.