Abstract:

The perception that an electronically generated sound source originates from outside the head at a particular direction can be produced by filtering the sound with head-related transfer functions (HRTFs). The perception that the source is at a particular distance is more difficult to produce, particularly the first time an unfamiliar sound is heard. In this study, listeners tried to identify source distances when a 300-ms white noise burst was filtered with transfer functions from a physically based atmospheric model including scattering by turbulent eddies in the atmosphere (``turbules``). Distances of 200, 400, 800, 1600, and 3200 m were simulated using a fixed distribution of turbules. Listeners' performance with and without inclusion of scattering by turbules was compared. Results indicate that: (1) when the subjects were given known sounds for reference or had become familiar with the test sounds, they correctly identified the distances about 80% of the time, despite changes in playback gain or reduced dynamic range, and (2) distance identification improved by about 10% when the effects of scattering were included. The improvement in distance identification was produced without the need for increased real-time computation over simple atmospheric absorption filtering. Larger improvements are expected with time-varying and spatial filters.