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

Re: Temporal encoding of pitch from resolved harmonics



Hi Chris, Annemarie,

Various groups of people use the terms spectral information
and spectral cues in different ways.

It helps to be clear about whether one is talking about
1) frequency-related structure in the power spectrum of the stimulus,
(acoustical sense) or
2) frequency-based neural representations of the stimulus,
(neurophysiological sense) or
3) the ability of listeners to discriminate different acoustic spectra
(psychophysical sense).

I know that it is an unwritten assumption amongst many psychophysicists
that
psychophysically-resolvable harmonics imply some kind of frequency-based neural
representation (usually some variant of a rate-place representation,
often sharpened
up by using neural synchrony and/or lateral inhibition).

That harmonics are resolved psychophysically does not necessarily imply
that rate-place profiles in
neural frequency maps (however one wants to construe these) will contain
the information
necessary to distinguish individual partials and to accurately estimate
their frequencies
(as one might do in a partial-matching task).

There are alternative possible neural correlates of resolvability in
neural interspike interval
distributions which involve the ability of (putative) central interval
analyzers to
distinguish and recognize multiple interval patterns in a composite
whole. For
example, make a series of harmonic complexes of F0=100, in the series
of 4, 4+5, 4+5+6+7,... or in the descending series 5, 5+4, 5+4+3, etc.,
and compute
the autocorrelation functions of the individual partials (the
autocorrelation of a sinusoid
is a cosine at the same frequency).

(For these kinds of low frequency
stimuli and rough demonstrations of principle, the autocorrelation is
not a terrible
zeroeth-order surrogate for an ANF population-interval distribution).

If one looks at the correlations between the autocorrelation
of the complex and the autocorrelation of the harmonic of interest (say
the 4th), the
correlation drops greatly with the addition of extra harmonics,
especially after 3 are added.
Visually, the pattern related to the harmonic of interest becomes harder
and harder to discern,
and the auditory counterpart would be that it is increasingly harder to
hear it out. It would be
interesting to determine whether resolvability (and the effects of
harmonic mistunings on fusion) can be
accounted for in terms of correlations between interval distributions.

Peter Cariani

Peter Cariani, Ph.D.
Eaton Peabody Laboratory of Auditory Physiology
Massachusetts Eye & Ear Infirmary
243 Charles St., Boston, MA 02114 USA

tel@EPL (617) 573-4243
tel@MGH (617) 726-5419
FAX (617) 720-4408

Email peter@epl.meei.harvard.edu
Web: www.cariani.com


Chris Plack wrote:
>
> Hi Annemarie,
>
> >
> >are there any studies reporting pitch extraction from low, resolved
> >harmonics that can only be attributed to temporal phase locking but not
> >to spectral encoding?
>
> Sounds like an impossible experiment!
>
> The very fact that the harmonics are resolved implies that there must be
> spectral information. You might be able to argue about whether the results
> are more consistent with a spectral or a temporal explanation (as many
> have), but I don't think you'd be able to rule out spectral cues entirely.
> At least not for a common-or-garden complex tone.
>
> I suppose you could construct the harmonics using bandpass AM noise at
> different rates or something, but that would be cheating!
>
> Cheers, Chris
>
> *************************************
>           Chris Plack - psychoacoustician, pop star
>        Department of Psychology, University of Essex,
>           Wivenhoe Park, Colchester, CO4 3SQ, UK.
>                       Tel: (01206) 873493
>                       Fax: (01206) 873590
>  http://privatewww.essex.ac.uk/~cplack/welcome.html
> *************************************