Effects of simulated reverberation on the use of binaural cues and fundamental-frequency differences for separating concurrent vowels

A computational simulation was used to generate impulse responses between points in a rectangular room and two points on opposite sides of a spherical “head”. Sounds were convolved with the impulse responses to generate stimuli with which to study the effects of reverberation on the ability of listeners to use differences in fundamental frequency (Δ Fos) to separate concurrent vowels. Experiment 1 verified the suitability of the simulation by showing that it produced (i) appropriate percepts of lateralization, (ii) a larger contribution to lateralization from interaural differences in timing than level, and (iii) no effects of reverberation on lateralization. Experiments 2–5 measured masked identification thresholds for synthetic harmonic “target” vowels in the presence of masking sounds. In Experiment 2, listeners identified targets against pink-noise maskers. The experiment established a spatial geometry and a degree of reverberation for which listeners did not benefit from binaural cues arising from the spatial geometry of the sources. Experiment 3 demonstrated that the same arrangement did not undermine the ability to use Δ Fos to separate targets from vowel-like maskers when both had static Fo contours, but did prevent listeners from using Δ Fos carried on coherently changing Fo contours. Experiment 4 showed that a modulation width of ±1.45% was sufficient to reduce the benefits of Δ Fos, but that the benefits were not eliminated until the width of modulation exceeded the Δ Fo. It is argued that these results are compatible with existing models of the ability to use Δ Fos to separate concurrent vowels and that reverberation undermines the ability when the Fos are changing by diffusing the periodicities of the competing sources. Finally, Experiment 5 demonstrated that reverberation had no effect on the ability to separate a modulated vowel from pink noise. Thus, reverberation may have its detrimental effects in these experiments by diffusing the periodicity of the masking sounds rather than the targets. Overall, the experiments demonstrate that Δ Fos can be more robust cues for separating concurrent sounds than binaural cues. The relevance of these results to the perception of natural continuous speech is discussed.