[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [AUDITORY] Localizing smoke detectors - why is it so hard?

Pierre, yes, those are good recent papers.  But the basic idea of ITD at high frequencies goes way back.  As McFadden and Pasanen (1976) say, based on bands of noise and two-tone signals, "In many conditions of listening, sensitivity to interaural time differences at high frequencies compares favorably with sensitivity at low frequencies--good performance requires only tens of microseconds of interaural time delay."

(McFadden, D. and Pasanen, E. G. (1976). "Lateralization at high frequencies based on interaural time differ-
ences", The Journal of the Acoustical Society of America 59, 634-639.)


ps. Their introduction lists older observations, and comments on how they tend to get ignored in the face of the duplex theory that works only with tones and narrowband signals; it is a good reminder for us now:

Recently Henning (1974a, 1974b) remade a point that
badly needed to be remade. He showed once again that
the auditory system can utilize interaural time differ-
ences to lateralize at high frequencies when the wave-
forms are complex. The same fact had previously
been demonstrated by David, Guttman, and van Bergeijk
(1959), Leakey, Sayers, and Cherry (1958), Klumpp
and Eady (1956), Tobias and Schubert (1959), and oth-
ers, but many of us concerned with binaural processing
had been ignoring these data and had come to think
solely in terms of the so-called duplex theory of sound
localization. This widely cited theory asserts that the
auditory system uses interaural time differences to
localize only for frequencies below about 1200-1500 Hz,
and that at higher frequencies it can only use the gradu-
ally increasing interaural level differences. But as
Henning (1974a, 1974b) showed, and as we shall further
demonstrate here, time-delayed complex waveforms at
high frequencies can be lateralized just as accurately
as time-deiayed low-frequency waveforms. The im-
plication is that with complex waveforms the auditory
system is able to process the ongoing interaural time
differences present in the fluctuating envelope. The
problem with the duplex theory, then, is that it was
based solely on localization and lateralization research
employing pure tones, and thus, is concerned only with
cycle-by-cycle differences in the fine structure of the
waveform. And while the theory is true enough for
such waveforms, it is almost totally irrelevant, for it
cannot be appropriately generalized to real-world situ-
atious involving complex waveforms.

On Thu, Jun 27, 2013 at 9:41 PM, Pierre Divenyi <pdivenyi@xxxxxxxxxxxxxxxxxx> wrote:
You are right, Dick! A series of recent papers on "transposed AM sounds" by Les Bernstein and Tino Trahiotis show very clearly that periodically interrupted high-frequency carriers are localizable based on ITD.

On 6/27/13 12:37 PM, "Richard F. Lyon" <dicklyon@xxxxxxx> wrote:

Ewan, thanks for your paper reference; very relevant.

You wrote there, "ITD dominance is shown indirectly in findings that head movements are highly effective for localizing low-frequency targets but not narrow-band high-frequency targets."

I agree.  But it doesn't address what you could do with wide-band high-frequency targets.  If the alarms used 3 kHz, but chopped on and off, with not such a high-Q resonance, they would probably have good enough onsets to help you make use of ITD, yes? 

I often see people disregarding ITD as a powerful cue above 1 kHz or so; but the basis for that is only that it's not a usable cue for sine waves and other narrow-band signals.  Wideband clicks and noises are easy to localize, even if simulated with only ITD.


On Wed, Jun 26, 2013 at 1:52 PM, Ewan A. Macpherson <ewan.macpherson@xxxxxxxxxx> wrote:
Richard F. Lyon wrote, On 6/25/2013 1:43 PM:


I believe the answer is primarily in the transducer:  to make the beeper
cheep, they use a resonant transducer, which has a slow buildup at the
onset and makes the resulting signal not very broadband at all,
depriving you of all ITD cues.  And they make the beeps so brief that
you don't have much chance to turn your head and vary the ILD cue;

It also turns out that front/back location is much more readily disambiguated by head turning in stimuli that carry low-frequency ITD than in those carrying only high-frequency ILD (such as the ~3-kHz, more-or-less pure tones from smoke detectors). The dynamic ILD cue does not seem to be able to beat the phantom spectral cue due to the narrow high-frequency peak in the spectrum. This is true under anechoic conditions, and presumably would be even worse in reverberation.