[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Is there considerable phase locking up to 6 kHz?
At 03:37 17/03/2004 -0500, you wrote:
I think this paper is relevant to this approach.
K. Krumbholz, R.D. Patterson, A. Nobbe, and H. Fastl. (2003). Microsecond
temporal resolution in monaural hearing without spectral cues? J. Acoust.
Soc. Am. 113 (5):2790-2800.
I agree that the highpass filtered conditions are highly relevant. We have
another earlier one that I think you will also find relevant.
Yost, W.A., Patterson, R.D. and Sheft, S. (1998). The role of the envelope
in processing iterated rippled noise. J. Acoust. Soc. Am. 104 2349-2361.
Again we show evidence for the preservation of time-interval processing
sufficient to make a 2AFC judgement when the stimuli are highpass filtered
at 6 kHz.
The place to look is in the Discussion. The relevant paragraphs are
I would be happy to send you a copy of the pdf if you would like.
We argue that the pitch difference between IRN stimuli generated with g=1.0
and those generated with g=-1.0 is based on waveform fine structure in
frequency regions that may be as high as 6000 Hz. Frequencies in the region
of 6000 Hz are above the region where phase locking is usually assumed to
operate. The auditory system would need to determine the temporal structure
of the IRN stimuli based on the delay (d) which is on the order of 4 to 16
ms. That is, the auditory system does not need to follow short temporal
events associated with the waveform fine structure of the high-frequency
IRN stimuli, but rather much longer duration regularities that are found in
the IRN stimuli. Nevertheless, the ability to process even these longer
temporal regularities in these very high-frequency channels appears
inconsistent with the use of a lowpass filter with a 1200-Hz cutoff as a
model of the loss of phase locking at high frequencies. However, such a
lowpass filter is only down 30 dB at 6000 Hz. Thus, it is possible that
some fine-structure information will "leak" through at these high
frequencies. Such fine-structure information might be made more clear if
cross-spectral processing is used. Figures 12 and 13 shows summary
autocorrelograms for an IRN stimulus with d=4 ms, 8 iterations, g=1.0 (Fig.
12) and =-1.0 (Fig. 13), and a 6000- to 8000-Hz filter. These summary
autocorrelograms were generated using the same modeling conditions used for
Figs. 4-6, including the 1200-Hz lowpass filter to simulate the loss of
phase locking. The summary correlogram is a cross-spectral processing
scheme representing the sum of the autocorrelation functions of the
individual channels across the 6000-8000 Hz region for each lag. As can be
seen there are differences in the summary autocorrelogram between the g=1.0
and g=-1.0 conditions that might be the basis for discriminating these two
stimuli. These differences in the summary autocorrelograms are much smaller
for the 8000- to 1000-Hz conditions. Thus, the long-duration temporal
structure that differentiates conditions when g=1.0 and from those when
g=-1.0 might still exist in the processed waveform fine structure even at
high frequencies, if something like a summary autocorrelogram was used as a
way to extract the information. Clearly a more quantitative model is
necessary to test the ability of such summary-autocorrelogram differences
to account for data like those shown in Figs. 7 and 9.In this paper, we
simply want to indicate that such differences in the summary
autocorrelograms do exist.
The work on IRN stimuli suggests that periodicities are not a necessary and
sufficient condition for processing complex pitch, since there are no
periodicities in the IRN waveform (see Yost et al, 1996), yet they have a
clear pitch (see Yost, 1996a). Our current work suggests that the
information for processing the pitch of IRN stimuli is not in the envelope.
That is, the envelope by itself cannot be used to explain the
discriminations between IRN stimuli generated with g=1.0 and g=-1.0. Thus,
models based on either periodicity processing and/or pure envelope
extraction would not be good ones for dealing with the pitch of IRN
stimuli1. For instance, neural channels tuned to periodic envelope
fluctuations would probably have a difficult time extracting the temporal
regularity that appears to be responsible for the pitch of IRN stimuli. Any
model of pitch based on envelope extraction would have to allow for
fine-structure information to be available in the frequency regions below
* ** *** * ** *** * ** *** * ** *** * ** *** * ** *** * ** *** * ** ***
Roy D. Patterson
Centre for the Neural Basis of Hearing
Physiology Department, University of Cambridge
Downing Street, Cambridge, CB2 3EG
phone 44 (1223) 333819 office
phone 44 (1223) 333837 lab
fax 44 (1223) 333840 department