Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!news.sprintlink.net!noc.netcom.net!netcom.com!sarala
From: sarala@netcom.com (Sarala Rajagopalan)
Subject: Speech Activity Detection
Message-ID: <saralaD9GIEB.DpB@netcom.com>
Organization: Netcom Online Communications Services (408-241-9760 login: guest)
Date: Wed, 31 May 1995 19:06:10 GMT
Lines: 20
Sender: sarala@netcom4.netcom.com

Hi,

I need to detect sections of speech in a digital recording (sampling 
frequency = 8 KHz) of single-speaker speech. The detection should
be robust enough to be unaffected by the presence of some background 
noise. 

I am considering exploiting the fact that speech is fast-changing. 
Sections (60 msec, for example) of a recording that exhibit high 
variation in either energy or number of zero-crossings would be 
classified as speech. Conversely, section of the recording 
that do not exhibit high variation in either energy or number 
of zero-crossings would be classified as not containing speech.

All comments/suggestions/pointers appreciated.

Sarala
**********************************************************************
Sarala Rajagopalan			sarala@netcom.com

