Newsgroups: comp.speech
Path: lyra.csx.cam.ac.uk!warwick!pipex!howland.reston.ans.net!europa.eng.gtefsd.com!MathWorks.Com!zombie.ncsc.mil!afterlife!jpcampb
From: jpcampb@afterlife.ncsc.mil (Joe Campbell)
Subject: Re: Question on PARCOR coefficients
Message-ID: <1994Apr9.204500.13663@afterlife.ncsc.mil>
Organization: The Great Beyond
References: <CnqJ2K.2ny@ecf.toronto.edu>
Date: Sat, 9 Apr 1994 20:45:00 GMT
Lines: 25

In article <CnqJ2K.2ny@ecf.toronto.edu>,
LIU  CHEUNG WAH STEPHEN <liuca@ecf.toronto.edu> wrote:
>	For the PARCOR coefficients used in Linear Predictive Coding,
>how should I re-calculate a coefficient that is unstable? (That
>is, the PARCOL coefficient is bigger than 1 or less than -1)
>	Thanks.

Three popular methods are to:

1) Reflect unstable poles about the unit circle in the z-plane.

2) Recompute the coefficients using windowed autocorrelation analysis
   that has guaranteed stability (beware of fixed point arithmetic!).

3) Repeat the previous set of coefficients, or, if you have delay,
   interpolate coefficients between adjacent stable frames (note,
   beware of the domain you interpolate in, e.g., interpolation of
   stable predictor coefficients doesn't guarantee a stable filter!)

Joe
-- 
.............................................................................
; Dr. Campbell  N3JBC  jpcampb@afterlife.ncsc.mil  74040.305@compuserve.com ;
; Speaking for myself     Happiness = Reality - Expectations, Click & Clack ;
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
