how the ear might "calcualte" sensory dissonance

One of the things that always bothered me about sensory
dissonance models was that they require a decomposition
of the sound into partials (say by an FFT) and then a step
where the psychoacoustic curves from Plomp and Levelt
(or similar researchers) is invoked. This is unplausible
physiologically, to say the least. Marc Leman (of the
Institute for Perception and Electronic Music) and I began
to think about alternate mechanisms, and came up with the
following idea, which we have just submitted to the Journal of
the Acoustical Society of America.

The basic idea is to use the same blocks that models
of pitch perception and masking use - and to then modify them
slightly so that they give the familiar dissonance curves.
Heres the title and abstract. If you'd like to see the paper, let me
know.

A Time-Based Model of Sensory Dissonance

This paper describes a time-based model for the calculation of
sensory dissonance that is similar to the `front end' of modern pitch
extraction and amplitude modulation detection models. The critical
band filters feed a nonlinear rectification unit; the output is bandpass
filtered and the energy is accumulated over a short time interval.
This gives the energy in the beating, which is a primary component
of the `roughness' or sensory dissonance percept. The model is
analyzed in a few simple cases, and a series of simulations
demonstrate that it successfully reproduces standard dissonance
curves. Since the model utilizes many of the building blocks of
standard auditory models, sensory dissonance can be viewed as a
byproduct (or co-product) of these neural elements. Amplitude
variations are handled automatically (unlike standard dissonance
models) and the relationship between sensory dissonance and the
critical band is a natural consequence of the form of the model.

On Wed, 17 Mar 1999, William Sethares wrote:
> One of the things that always bothered me about sensory
> dissonance models was that they require a decomposition
> of the sound into partials (say by an FFT) and then a step
> where the psychoacoustic curves from Plomp and Levelt
> (or similar researchers) is invoked. [snip]
>
> The basic idea is to use the same blocks that models
> of pitch perception and masking use - and to then modify them
> slightly so that they give the familiar dissonance curves.
> Heres the title and abstract. If you'd like to see the paper, let me
> know.
[snip]

This sounds great, Bill! I'm surprised no one else has commented. I'd
love to see the paper; I just hope I can follow it.

--pH <manynote@lib-rary.wustl.edu> http://library.wustl.edu/~manynote
O
/\ "How about that? The guy can't run six balls,
-\-\-- o and they make him president."

NOTE: dehyphenate node to remove spamblock. <*>