From: malcolm@apple.com (Malcolm Slaney)
Date: Thu, 28 Jul 94 09:01:43 PDT
Subject: Sound Separation at CCRMA
Message-Id: <9407281601.AA05672@federal-excess.apple.com>

We're back!  The Hearing Seminar returns to Stanford CCRMA with a
distinguished visitor from Japan talking about his sound separation work.
Dr. Hiroshi Okuno (NTT) will be in town at the end of next week and will
talk about his work on agents for sound separation.   Hurray!

A number of sound separation systems have been described at CCRMA.  We've
talked about the cues that are used for segragating sounds, but how do we
put them back together again?  How do we decide which components belong
together and build software to make the decisions?  Dr. Okuno's talk will
address these issues.

        Who:    Hiroshi Okuno (NTT)
        What:   Multi-Agent System for Auditory Stream Segregation
        When:   Friday August 5 at 2PM          <<<<**** Note Different TIME
        Where:  CCRMA Library (Top Floor of the Knoll at Stanford)

I'm also working on the schedule for the rest of the year.  Dan Ellis (MIT
Media Lab) will talk about his sound separation work later in August.  Who
has been doing neat work, or read an interesting paper that would make a
good discussion at CCRMA?  Please send me your plans, ideas, and
suggestions.  Remember, this is a group effort and we all benefit when we
can share ideas.

Note next week's talk is at a different time (Friday 2PM).  Unless there is
a concensus against it, we'll return to our normal 10:30 time on Thursday
mornings.

I've attached an abstract for Dr. Okuno's talk and a bibliography to the
end of this note.  See you at CCRMA

-- Malcolm

Auditory Stream Segregation in Auditory Scene Analysis
         with a Multi-Agent System

Hiroshi G. Okuno  with Tomohiro Nakatani and Takeshi Kawabata

(NTT Basic Research Laboratories)

Abstract

In this talk, I first discuss the motivations of auditory
scene analysis research from the viewpoint of computer
science.  Then, I propose a novel approach to auditory
stream segregation which extracts individual sounds ({\it
auditory stream}) from a mixture of sounds.  The HBSS ({\it
Harmonic-Based Stream Segregation}) system is designed and
developed by employing a multi-agent system.  HBSS uses only
harmonics as a clue to segregation and extracts auditory
streams incrementally.  HBSS has two kinds of agents, tracer
and tracer-generator.  When the tracer-generator agent
detects a new sound, it spawns a tracer agent, which
extracts an auditory stream by tracing its harmonic
structure.  The tracer sends a feedforward signal so that
the generator and other tracers should not work on the same
stream that is being traced.  The quality of segregation may
be poor due to redundant and ghost tracers.  HBSS copes with
this problem by introducing monitor agents, which detect and
eliminate redundant and ghost tracers.  HBSS can segregate
two streams from a mixture of man's and woman's speech.  It
is easy to resynthesize speech or sounds from the
corresponding streams.  Additionally, HBSS can be easily
extended by adding agents of a new capability.  HBSS can be
considered as the first step to computational auditory scene
analysis.  Demonstration will be performed after presentation.

This talk is based on our paper of AAAI-94, which is available
by anonymous FTP:
       sail.stanford.edu:/pub/wj/aaai94.ps


Bio

Hiroshi G. Okuno is a staff engineer of NTT Basic Research
Laboratories.  He received BA from Department of Pure and
Applied Science, the Unviersity of Tokyo, in 1972.  He
joined NTT Basic Research Labs, designed and developed Lisp
systems and Lisp machine for more than ten years.  He
started a research of a multi-agent system and auditory
scene analysis around 1991 and organized a tiny group in
1992.  He stayed at Knowledge Systems Laboratory, Stanford
University, as visiting scholar from 1986 to 1988 and was
engaged in parallel processing for artificial intelligence.
He received the 1990 best paper award from Japanese
Association for Artificial Intelligence.  He was visiting
associate professor of endowed chair of intelligent
engineering, the University of Tokyo, from 1992 to 1993.  He
is also a part time lecture at several universities.  He is
on the board of directors of Japanese Society for Software
Science and Technology.
-------------------------------------------------------------