Re: Granular synthesis and auditory segmentation (at)

Subject: Re: Granular synthesis and auditory segmentation
Date:    Fri, 23 Oct 1998 08:04:19 +0200

Jont Allen asked > Are two tones that beat stationary or non-stationary, > by your definition? Stationary. The frequency content is the same at all times in your example. This is how stationary is defined in signal theory. > Now there are two reasons that the demodulated (neural) signal > cannot have components above 5 kHz. First is the well known 4 kHz > lowpass filtering of the synapse, that causes the loss of pure tone > synchrony (Kiang monograph + Don Johnson PhD Thesis), and the second > is the cochlear filter bandwidth which restricts the bandwidth of > the rectified neural resp. That is why my hypothesis was about the 3-5 kHz range. Then you came up with the 10 kHz beat example that had a 50 Hz demodulated neural signal. Neither of these two figures were in the 3-5 kHz range I was inquiring about. Like I said before, what I basically want to know is what the "volley principle" nerve frequencies in the 3-5 kHz range bring us functionally. > In my opinion, even with the "volley principle" you cant process > signals that have been removed by a filter, as in this case. The > volley principle cant improve the SNR and recover the signal, since > No processing can do that. If the filter has removed something, you don't get it back, of course. My point is that signals in the 3-5 kHz range do turn up in the "volley principle" statistics (at least in the cat), and in that case neural processing *might* do something with that, or can you prove that that is impossible? Can you prove that "volley principle" + subsequent neural processing cannot increase SNR and recover the signal? > As I said above, you cannot recover signals that have been removed > to less than the noise floor by filtering with the volley principle. Why not? If the signal is a pure tone there will be long term autocorrelations in the signal that you can in principle use to detect the signal whatever the SNR. This is how deep space probes like Voyager are tracked. Of course, required signal integration times move up when SNR goes down, leading to various practical limitations. > I guess I am agreeing with your first statement, if modified such > that you replace the 3-5 kHz range to say "above the frequency > of the loss of pure-tone neural syncrony (i.e., 4 kHz): Actually my 3-5 kHz is your "4 kHz" plus a fair 1 kHz margin. Single-neuron synchrony will have broken down well below 3 kHz, and the "volley principle" is what creates synchrony in the 3-5 kHz range. Not necessarily right up to 5 kHz, but certainly above 3 kHz (at least in cats), so within the 3-5 kHz range. > > With the reformulated request I tend to maintain that the > > place theory of hearing is sufficient to account for what > > is observed psychophysically in the 3-5 kHz range, rather > > than say that it is the only possible account. As a matter > > of fact, temporal processing could in principle account for > > everything, since it can encompass any type of filterbank. > > However, if it really were that powerful, we would have no > > need for a cochlea at all, and evolution would have had > > little incentive to give us a cochlea. > > Now just wait a darn minute here. The main purpose of the > filter band is to clean up the signals before being nonlinearly > transduced. There are also major dynamic range issues that are > handled in the cochlea. You are taking a very small view of what > the auditory system is trying to do, and it cannot do it well, > in high noise situations, without a working cochlea. I would > be happy to discuss/argue this point with anybody. Like I said, temporal processing could *in principle* account for everything. I know that it wouldn't work well in practice with a purely neural temporal processing implementation, for instance because of dynamic range issues. So your argument just confirms what I said, because I was arguing for cochlear filterbank processing within the 3-5 kHz range, in line with my hypothesis. Maybe the confusion arose because to me "place theory" in the 3-5 kHz range includes the subsequent neural processing up to, say, 1 kHz to account for things like beat detection. Thus, a 4 kHz input signal would in that view only be detected via "place theory", unless supplemented by something based on the "volley principle". I am still looking for clear counterexamples demonstrating (or making plausible) that "volley principle" interspike periodicities in the 3-5 kHz range lead to psychophysically observable effects in humans. Peter Cariani so far got closest with his octave matching case, but the outcome was not yet conclusive w.r.t. the value of the transition frequency (below or above 3 kHz). Best wishes, Peter Meijer Soundscapes from The vOICe - Seeing with your Ears! McGill is running a new version of LISTSERV (1.8d on Windows NT). Information is available on the WEB at

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University