Auditory target detection in reverberation

Abstract

Measurements and theoretical predictions of auditory target detection in simulated reverberant conditions are reported. The target signals were pulsed 13-octave bands of noise and the masker signal was a continuous wideband noise. Target and masker signals were passed through a software simulation of a reverberant room with a rigid sphere modeling a listener’s head. The location of the target was fixed while the location of the masker was varied in the simulated room. Degree of reverberation was controlled by varying the uniform acoustic absorption of the simulated room’s surfaces. The resulting target and masker signals were presented to the listeners over headphones in monaural-left, monaural-right, or binaural listening modes. Changes in detection performance in the monaural listening modes were largely predictable from the changes in target-to-masker ratio in the target band, but with a few dB of extra masking in reverberation. Binaural detection performance was generally well predicted by applying Durlach’s [in Foundations of Modern Auditory Theory (Academic, New York, 1972)] equalization-cancellation theory to the direct-plus-reverberant ear signals. Predictions in all cases were based on a statistical description of room acoustics and on acoustic diffraction by a sphere. The success of these detection models in the present well-controlled reverberant conditions suggests that they can be used to incorporate listening mode and source location as factors in speech-intelligibility predictions.