The most obvious clue to the distance of a sounding object is its amplitude. It has been shown that for a constant point source in a free sound field, the sound pressure level falls of by six decibels with every doubling of the distance [Blauert, 1997]. However, amplitude can only serve as an absolute measure of distance if the listener is familiar with the source signal as the amplitude of an unfamiliar or synthesized sound can only provide a relative estimate of a change in distance [Mershon, 1997; Sheeline, 1983]. For example, it is easy to determine whether someone is speaking quietly nearby or shouting loudly far away because of the
familiar and characteristic dynamics of the human voice, but when the source signal is unfamiliar this distinction cannot be made. Various studies under free-field
conditions have found that for a source with constant amplitude, there is no
relationship between the actual and perceived distance [Nielsen, 1993; Mershon et al, 1975], indicating that, in this case, the signal amplitude can only indicate how the relative distance is changing. For very close sources (< 1m), it appears that the auditory system also uses additional binaural localization cues to estimate the source distance. Brungart analysed a number of HRTF measurements which showed that very near sources result in substantial changes in ILD [Brungart et al, 1999], although the ITD remains largely unchanged.
It has long been known that the perception of distance is also influenced by the effect of acoustic reflections in the listening environment. The level of the direct and early reflected sound will change substantially as the distance from the source to the listener changes. However, the level of diffuse reverberation is largely independent of the position of the listener in the room. Therefore, as the source distance increases, the direct sound will decrease while the reverberant sound remains constant. Beyond a certain distance, the reverberant signal level will be greater than the direct signal level, and the perceived distance becomes fixed and independent of the actual source distance. This critical distance is indicated in Figure 2.5 [Howard et al, 1996]. Various studies have shown that in reverberant rooms, the perceived distance of a real source is independent of the source level [Nielsen, 1993], which suggests that the ratio between the direct and reverberant signals, the D/R ratio, is a significant distance cue in real rooms. This theory was first proposed in the 1960s and this simple ratio is still commonly used by sound engineers and producers to control the depth of
different sources in two-channel stereo mixes.
Fig. 2.5 Source distance v sound intensity
The D/R ratio can provide a relative sense of distance but this simple ratio ignores the fine spatial and temporal structure of the reflected indirect signals. Experiments by Kendall reported that a strong impression of distance was perceived when listening to dry test signals augmented solely with a limited number of artificial early reflections, even when these reflections were restricted to those which followed
similar test which also found that a better distinction of distance was achieved when simulated early reflections were added instead of solely diffuse reverberation [Michelsen et al, 1997]. Neher investigated the perceptual effect of different early reflection patterns and found that listeners were unable to distinguish between an early reflection pattern comprised of accurately panned reflections, and one that was physically identical except that each reflection was simply reproduced by the nearest available loudspeaker. This suggests that although spatial differences in the early reflections pattern are perceptually salient, the actual angles of incidence of reflections may not be crucial [Neher, 2004].
Michael Gerzon presented a similar model of distance hearing, based on a theory originally proposed by Peter Craven [Gerzon, 1992b]. The Craven hypothesis assumes that the apparent distance of sounds is derived from the relationship between the relative time delay and amplitude of the early reflections and the direct signal. Gerzon and others have suggested that closely-spaced or coincident microphone techniques having a substantially omnidirectional total energy response will reproduce the absolute source distance better than microphones with a more directional response [Gerzon, 1992b; Theile, 1991]. Gerzon also points out that although it is now known that the simple direct/reverberant ratio does not provide an absolute measure of distance, it is still a useful subsidiary cue for relative distance, and is thus preferably made consistent with the apparent distance [Gerzon, 1992b].
In general it has been found that the perceived distance of a sound source in a room is compressed, as it increases virtually linearly with source distance at short range, but converges to a certain limit when the source distance is increased beyond the critical distance [Mershon et al, 1975; Nielsen, 1993]. There is therefore a non- linear relationship between the perceived and actual source distance. Bronkhurst suggests that this non-linearity arises not only due to the different mechanisms involved for different source distances, but also because the auditory system is not always able to accurately separate the direct and reflected signals [Bronkhurst, 2002]. In a number of listening tests, the effect of early reflections on perceived source distance was assessed in terms of the angle of incidence of the reflected sound. When the lateral walls were made completely absorbent, the source was perceived to be close to the head, virtually independently of the actual source distance. When lateral reflections were introduced, the perceived distance more closely matched the actual
2002]. This suggests that the direct to reverberant ratio is estimated by the auditory system using directional binaural cues to separate the direct and reverberant signals, although further tests are needed to confirm the validity of this hypothesis.
While clearly the amplitude of the source signal and the specific relationship between the direct and indirect signals have been shown to be dominant cues in distance perception, other secondary cues also provide some indication of relative distance. In general, an increase in source distance results in a reduction of the high frequency spectral content of the source signal due to the effect of air absorption. This occurs at large distances outdoors, but also in rooms due the absorptive nature of the boundary surfaces and the large overall distances travelled by the indirect signals as they reflect around the room.
The perceived shift in frequency due to the movement of a source relative to the listener, i.e. the Doppler effect, also provides an indication of the relative motion of the source.
2.3.1 Summary of Spatial Hearing
The preceding section summarized the different localization mechanisms which are used by the auditory system to determine the direction and distance of a source signal. It is still unclear, however, as to how these multiple, different and potentially conflicting localization cues are resolved into a single distinct spatial impression. Most theories of auditory localization now propose that the perceived source location is the one that satisfies as many of the localization cues as possible. When multiple conflicting localization cues are present, a simple majority decision is used to determine the location of the sounding object. Most real sources produce a complex signal which will produce localization cues that support each other in the frequency ranges at which they dominate. In addition, the corresponding visual component of a real source will also support the direction suggested by the auditory senses. Broadband signals are able to satisfy more localization mechanisms than narrowband signals and are therefore, in general, easier to locate. In addition the presence of onset transients greatly increases localization accuracy. Narrowband signals such as sine tones, particularly in the 2kHz region, are difficult to localize even under ideal conditions as no localization mechanism is particularly effective in this region. Early reflections and reverberation can also have a significant effect on
the perceived size of the source, which can result in a corresponding decrease in localization accuracy.