Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!uknet!pipex!uunet!cs.utexas.edu!sdd.hp.com!nigel.msen.com!heifetz!agog!greg
From: greg@agog.com (Greg Cockroft)
Subject: Re: LPC synthesis. Help.
Message-ID: <1993Aug4.233751.1010@agog.com>
Keywords: LPC synthesis
Sender: greg@agog.com (Greg Cockroft)
Organization: Agog, Inc.
References: <CB6t7K.HMw@ccu.umanitoba.ca>
Date: Wed, 4 Aug 1993 23:37:51 GMT
Lines: 26

In article <CB6t7K.HMw@ccu.umanitoba.ca> gouri@eeserv.ee.umanitoba.ca  
(SriGouri Kamarsu) writes:
> I am working on LPC synthesis and I need some help.
> I have a speech data set. I do lpc analysis with a window size of 320  
samples
> and a 50% overlap. When I do my synthesis, how do I take care of the
> overlapped synthetic speech? If the question is not clear, 

You do not want an overlap. The only reason you have an overlap for the  
computation is that you need enough data points to make sure you end up  
with a stable system. 

On synthesis for voiced speech you will want to update the lpc coffs at  
the beginning of each pitch period. The easiest way to do this is to pick  
the closest set of lpc parameters in time. For 8khz data 80 samples would  
be a reasonable pitch period. You could use each of your lpc parameters  
for 2 pitch periods. 


	-greg.

-- 
Greg Cockroft                       greg@agog.com    
Agog, Inc.                          (517) 627-2186   fax 2673
13780 North River Highway
Grand Ledge, MI  48837
