Echo suppression in the horizontal and median sagittal planes

Abstract

Experiments were performed to measure two kinds of suppression threshold for running speech: echo threshold, defined here as the minimum level at which it was possible to detect that an echo was present, and masked threshold, defined as the minimum level at which it was possible to detect that a lagging sound was present at all. Both thresholds were measured using a geometry in which sound sources and reflections were distributed over the horizontal plane (left, front, and right locations) and a geometry in which they were distributed over the median sagittal plane (front, overhead, and rear locations). The predominant sound localization cues are different for these two geometries, and the experiments measured the consequences of this difference for suppression. Echo thresholds were found to have a comparable dependence on the delay of the lagging sound for the horizontal and median sagittal planes. Masked thresholds, which were systematically 8–15 dB lower than echo thresholds, also showed a comparable dependence on delay for the two planes. Overall, these results support the idea that echo suppression is functionally similar whether locations are cued by interaural differences in time and intensity, or by spectral features introduced by the head-related transfer function.