Forums

On G.729 Perceptual weighting and Gain Quantisation

Started by Thenral p November 9, 2001
Hi all,
I am working on G.729 and when going thru' the C code I came across the
folowing:
In perceptual weighting filter to find the value of gamma2(weighting
parameter) the formula used is
gamma2=1-6.0*pi*dmin

whereas in the G.729 code it is given as gamma2=1-6.0*dmin

I would like to know which is the correct equation?
2)If a detailed description of the preselection procedure for gain
quantisation is available online can someone help out with the web site?

Thanks a lot
Thenral




From: "Thenral p" <> wrote:

>I am working on G.729 and when going thru' the C code I came across the
>folowing:
>In perceptual weighting filter to find the value of gamma2(weighting
>parameter) the formula used is
>gamma2=1-6.0*pi*dmin
>
>whereas in the G.729 code it is given as gamma2=1-6.0*dmin
>
>I would like to know which is the correct equation?

This rings a faint bell (from quite a few years back). The
value dmin is derived from line-spectrum frequencies. I remember
that when I wrote a floating-point g729 and compared it with
the official fixed-point code, there were confusions to do with
whether the LSF are considered to run from -1 to 1 or -pi to pi.

Probably you will find that both the equation and code are
correct, but use a different definition. In any case, if you
want to write a g729 the C code is definitive, however silly.

>2)If a detailed description of the preselection procedure for gain
>quantisation is available online can someone help out with the web site?

Not that I know of, I'm afraid.

Tim.
------------------------
Tim Thorpe Tel: +44 1223 240366
Fax: +44 1223 414402
------------------------