(Pierre Divenyi)

From:    Pierre Divenyi <marva4!pdivenyi(at)ucdavis.edu >
Date:    Tue, 8 Dec 92 15:24:55 PST

Dan Ellis's message is very thought-provoking. I wish that those of us interested could have a little glimpse at the algorithms Dan and David Huron (and others?) have been playing with and later discarded. Even failures are revealing, often more so than successes. Re/ the acoustic/auditory stream debate: It is without question that there ARE acoustic streams -- because there ARE multiple sources that decided to become active all at the same time and create a hell of an acoustic mess... This mess can be broken up (heuristically) into components using several methods. When we say "auditory", we refer to the method the human listener employs and we know that he is fairly successful. Now, there could be algorithms arriving at a similar success rate, using schemes conceivably very different from what, according to our investigations and speculations, human listeners use. Thus, "auditory stream" designates one solution to the problem of breaking up the acoustic scene into acoustic streams. Our attempts to either cast the listener's auditory-stream strategy into algorithms, or to design our own, also contribute to the enrichment of our vocabulary of scene analysis, of which we are intimately familiar only with the "auditory" family -- and even that only through experience (because we have not been able, as of yet, to formally describe this experience). Lots of luck to Dan, David Huron, and all of us -- it's an interesting game! Pierre Divenyi

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