19 th INTERNATIONAL CONGRESS ON ACOUSTICS MADRID , 2-7 SEPTEMBER 2007 Simulating distance cues in virtual reverberant environments

A series of behavioral experiments examined what cues listeners use to judge distance of nearby sources in reverberant spaces. Results show that, in real environments, reverberation aids performance even for sources within a meter of the listeners, where reverberant energy is relatively low compared to the direct sound. When simulating the same conditions using binaural head-related impulse responses, 1) performance in an anechoic environment is at chance level, while performance in reverberation matches real-room data, suggesting that reverberation makes simulated distances more robust; 2) monaural and binaural stimulus presentations produce similar results, indicating that binaural cues are less important for distance perception in reverberant environments. Finally, listeners cannot judge the distance of high-frequency medial stimuli with accuracy. For these stimuli, overall level is the only available distance cue, so distance judgments will only be reliable if distal stimulus level is constant. INTRODUCTION For sound sources within reach of a listener, interaural level differences (ILD) vary with both the direction and distance of the sources [1, 2]. This cue is absolute in the sense that the mapping from ILD to distance does not change with source content or environment, unlike relative cues such as the overall level of the received stimulus. Previous localization experiments performed in anechoic space suggest that listeners use ILD cues to judge source distance [3]. In reverberant environments, an additional distance cue is available, related to the direct-toreverberant energy ratio ([D/R, [4]) although D/R per se may not be the cue used by listeners [5]. Distance information in the D/R may be relatively weak for sources near the listener, because relatively little reverberant energy is received at the ears. Experiment 1 compares distance perception for nearby sources in anechoic and reverberant environments. Virtual acoustics techniques can be used to simulate distance of sound sources and to achieve precise stimulus control [6]. The remaining two experiments use this technique to manipulate available cues when judging distance. In Experiment 2, monaural and binaural performance is measured in anechoic and reverberant environments. Results are compared to those from the real environment to determine the relative contributions of ILD and reverberation cues. Experiment 3 looks at how performance depends on the center frequency of the stimulus. EXPERIMENT 1: REAL ENVIRONMENTS ANECHOIC VS. REVERBERANT SPACES Distance judgments in reverberation were collected and compared to anechoic data from a previous study using similar methods [3]. Because the reverberation has a relatively low intensity compared to the direct sound for sources near the listener, we hypothesized that performance would be similar in the two environments. Methods Seven subjects participated in Exp. 1, which was performed in a small classroom. The stimulus was a train of five 150-ms 200-8000-Hz pink-noise bursts with 30-ms gaps, presented from a small speaker placed at random locations in the listener’s right hemifield, within 1 m of his/her head. For each trial the presentation level was first adjusted so that the level at the center of the head was approximately constant and then roved by ±5 dB to eliminate overall level at the ears as a cue. Listeners indicated the perceived source location by pointing with a wand. Electromagnetic trackers were attached to both the sound source and the pointing wand, recording the source and response locations. Data were binned by the cone-of-confusion angle (i.e., the