neural network for timbre recognition

[Luis Felipe de Oliveira <luisfol@xxxxxxxxxx>]

Dear list,
We're getting some problems in creating the patterns for a network.
We're concerned with sound patterns recognition, and in this case, we need,
first of all, convert some sound files (like .wav or .aiff) in a matrix,
that will be feed in our netwotk. Our goal is to build a net for timbristic
recognition.
As far as we know, PCA or Karhunen-Loeve Transformed is the most suitable
mathematical tool for such a procedure, because it outputs seems to require
fewer neurons in the input layer.
Someone knows or have some experience in this kind of project?
How do you create the input matrix? with what program?
thanks a lot.
Luis Felipe Oliveira and Juliano Broli

-- 
Post-graduation students in Cognitive Science and Philosophy of Mind at São
Paulo State University;
Graduated in Composition at São Paulo State University.
Marília - SP - Brazil
•