Re: Can a timbre affected by a shifted virtual pitch evoked by double-spaced harmonics?

[Matt Flax <flatmax@xxxxxxxx>]

On Wed, Feb 19, 2003 at 03:34:44PM +0800, Chen-gia TSAI wrote:
> Hello everyone,
>
> My study of the timbre of the Chinese flute leads me to find a strange
> effect In my experiment, all complex tones contain five components with
> frequencies
...SNIP...
You should read this paper :
Resiude pitch as a function of component spacing
     Order Roy D. Patterson and Frederic L. Wightman J. Acoust. Soc. Am.
   59, 1450 (1976)

On Thu, Feb 20, 2003 at 05:11:23AM +1000, Harvey Holmes wrote:
> I believe that your observations may be well explained by Terhardt's
> virtual pitch theory, which is explained at
> http://www.mmk.ei.tum.de/persons/ter.html (go to the Virtual Pitch
> link).  Many related pitch references are also given on this site,
> including definitions of several different pitch concepts.
...SNIP...
I agree with Harvey,
in fact for the first two of your examples :
{1344 1844 2344 2844 3344}Hz yeilds a virtual pitch of 461Hz
{1680 2180 2680 3180 3680}Hz yeilds a virtual pitch of 545Hz
These are not the only virtual pitches, however the most salient
according to Terehardt's model.

On Thu, Feb 20, 2003 at 11:28:50AM +0100, Martin Braun wrote:
> This theory does not explain any pitch. It is anatomically and
> physiologically unrealistic, and the speed of f0 perception excludes
> any model of spectral pattern matching:
...SNIP...
Terhardt's models are not accurate due to anatomical modeling ...
They are accurate as they are an expert system. From what I see and
understand a very good one too. A large ammount of our modern music
compression systems use similar models which have been surely influenced
by Terhardt's work. Particularly masking.

On Sat, Feb 22, 2003 at 04:31:10PM +0800, Chen-gia TSAI wrote:
> This success of autocorrelation in modeling pitch extraction is, at
> least for me, very impressive. I also noticed that the autocorrelation
> function has been used to analyze musical timbres in
...SNIP...
Autocorrelation dates back to an article by Licklider (1951 in
Experientia). The concept that autocorrelation correctly models the
complex neural processing which is anatomically present in the inner ear
is a big question mark in my opinion. The beginning of my views can be
found here (this is only a place model- a temporal model is in the pipe
line): http://mffmpitch.sourceforge.net/

...SNIP...
> This approach appears to differ from the spectral analysis. My first
> question is: how far has the autocorrelation function been used in
> timbre research?
Yes it differs in that it is a temporal model ... which works in the
time domain
'spectral models' are termed place models ... they work in the
'spectral' domainIn general autocorreltaion was originally seen to work
in the afferent system in the temporal domain ... I don't disagree ...
however the systems are more complex then originally thought.

As far as timbre goes, I don't think I am qualified to talk about that
just yet.

On Sat, Feb 22, 2003 at 08:01:01PM +0100, Martin Braun wrote:
> As to the pitch model of autocorrelation, also this is a model that is
> anatomically and physiologically unrealistic. Ray Meddis, one of the
> major advocates of this model has now given it up, in favor of a new model
> (see below) that is based on anatomical and physiological data that were
> described in detail by Gerald Langner and me.
...SNIP...
I agree, however alot of the inner ear is unmeasurable using todays 'off
the shelf' technology. For example consider the efferent LOC returns.
It would be great to get electronic copies of these articles you
mention.

Matt

--
http://mffm.darktech.org

WSOLA TimeScale Audio Mod  : http://mffmtimescale.sourceforge.net/
FFTw C++                   : http://mffmfftwrapper.sourceforge.net/
Vector Bass                : http://mffmvectorbass.sourceforge.net/
Multimedia Time Code       : http://mffmtimecode.sourceforge.net/