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

Re: Is there considerable phase locking up to 6 kHz?



At 10:40 17.03.2004 +0000, jan schnupp wrote:
>Richard F. Lyon wrote:
>
>> Now that I've read Eckard Blumschein's note about how maybe "Pitch
>> apparently relates to a kind of autocorrelation that is not based on
>> phase locking,"
>>
>Richard is of course absolutely correct about this.

Of course. Being uncertain about the original meaning of the term phase
locking, I looked into Warren (1999) and found 'phase locking of spikes to
the stimulus waveform'. I would rather prefer 'to the result of modified
FCT and rectification' because the inner hair cells do not immediately see
the stimulus.

>I'm not sure whether that helps us with
>Cheng-Gia's original problem though, as, ...

A={400*9, 400*11, 400*13} Hz
B={400*11, 400*13, 400*15} Hz
C={400*13, 400*15, 400*17} Hz

Yes, if A, B, and C were played in succession, stimulated place on cochlear
partition was shifted by 400Hz. It didn't matter that the components were
unresolved. Tonotopic place of belonging T-multipolars - and as I imagine,
corresponding chopper frequencies - also changed. If we did actually hear
by frequency, then I would expect narrow-band tones without any ambiguity.
That's why I consider Chen-Gia's multi-pitches one more indication for
hearing joint ACF rather than frequency. The ACFs are presumably pretty
ambiguous in this case.

Let me explain why I see the spectral code a clever vehicle of temporal
structure:
- 'Above about 1,400 Hz, a single fiber cannot respond to successive
statements of a repeated waveform' and '...variability becomes so great for
a sinusoid near 4,000 Hz (period of 250 usec) that the temporal smear
obscures phase locking' (Warren).
- This is reflected in dependency of synchrony on frequency. Synchrony for
primarylike neurons is about the same as for AN fibers. These cells are
known to project to MOS for localization. Pitch is most likely conveyed via
T-multipolar choppers constituting the other large population of neurons
within CN. Here synchrony already drops at lower frequency, perhaps due to
down-sampling. Nonetheless, these choppers are firing with high temporal
precision.
- FCT is the only realistic cochlear transform, and it also fits best to a
second neural analysis. Unfortunately, Wiener-Chintchine theorem is not
well known among physiologists. Performing FCT twice, with rectification
and downsampling of frequency in between, returns something like ACF or
cepstrum, i.e. a signal in time domain being ready for further processing
and comparison. A simple frequency map would not be of much use for
recognition.
- While resolution of spectral shape is obviously rather poor, temporal
coincidence detection can much better account for the astonishing accuracy
of hearing. If we did hear by frequency, width of CB would waste accuracy.
It makes more sense to widen spectral bands in order to sharpen the final
temporal pattern.

Eckard