(back to
SpatialLocalizationPaper)
As alluded to previously, the ability of the auditory system to resolve complex sources accurately is not perfect. However, the most prominent effect -- the fusing of similar sounds that come from different directions -- actually helps in most cases by subconciously filtering reflections that would otherwise hinder speech perception. In particular, when the time difference between two such sounds is between 1 ms and 35 ms, the usually monaural & binaural directional characteristics of the first sound predominate, giving rise to the precedence or
Haas effect. A striking example shows a person standing in front of a horizontal row of speakers. Wherever he moves, the collective sound seems to always come from the speaker in front of him (Handel, 106).
When the sounds are less than 1 ms apart, the fused image lies between them instead of "preferring" the closer one. Above 35ms the sounds begin to separate into distinct auditory events; this threshold may actually be as low as 5ms for clicks (Moore, 162). (Given these requisite time delays, we must assume that the speakers in the previous example are at least 6 inches and no more than 25 feet apart.) The Haas effect becomes inconsistent with sounds located in the median plane (Blauert, 222). It also fails when the sounds are not transient. Gulick:
When a pulsed sinuosidal [sic] stimulus was split into two signals, one composed of the onset and offset transients and the other of the steady state, then when each of the signals was sent to a different loudspeaker, listeners located the sound at the transient speaker even when the signal from the steady-state speaker was of several seconds' duration. (340)
Finally, it has a dependence on the spectral similarity of the signals. Sufficiently similar sounds will fuse when the time delay is appropriate, with the perception becoming wider and more diffuse as the sounds differ (Handel, 107). The details of this effect appear to hinge on context, such that the audibility of distinct echoes depends on things as varied as the listener's experience, his visual stimuli, and his mental state (Blauert, 412).
Of course, even when not listening for echoes per se, a reverberent cathedral is clearly experienced differently from an anechoic chamber, with plenty of room in between. An interesting factor in rooms designed for listening is not only the total amount of echo, but the ratio of those which come from above or below and those which come from the side.* Since the bulk of our
BinauralLocalization occurs in the horizontal plane (pinnae elevation cues excepted), the former signals provide additional localization information (their front/back geometry being preserved) about the source while the latter would on their own suggest phantom sources to the left and right. Nevertheless, listeners tend to prefer halls with more lateral, image-diffusing echo. Architects can therefore please them using the precedence effect: by making the ceilings sufficiently high, the lateral reflections arrive first -- not so early as to be masked by direct sound, but close enough to mask the vertical reflections.
*Blauert gives extensive mathematical treatment of this principle, as always. See pp. 348-357.
onward to
StereoRecording
There are no comments on this page. [Add comment]