[ghost-dev] Quantization of sine params

[Jean-Marc Valin]

Hi Anup,

As you pointed out, the parameters c and d can (though I've never tested
that) be used to improve the frequency estimate as you suggested. That's
actually the main reason they're there. If that doesn't work, then we
might as well get rid of them and go with non-modulated sinusoids
(there's already code for that as well).

	Jean-Marc

On Fri, 2006-07-07 at 10:29 +0530, Anup K C wrote:
> Hi Jean-Marc
> 
> I am facing a few problems while trying out the harmonic grouping
> method. 
> 
> The decoder is expected to reconstruct the output from  w, ai,bi,ci
> and di vectors. So these are the parameters which are to quantized. I
> was expeceting that the w vector would contain the frequencies of the
> signal accurately or atleast it will  haev teh same characteristics as
> the  actual frequencies. But what I observe is that, thought the
> synthesised signal contains exactly same sines as teh actual signal
> the frequencies as indicated by w do not necessarily show the corerct
> values. And the biggest problem is that teh harmonic character of the
> sine waves is also not observable in the w vector. And I feel that
> this might make teh qunatization very tough; even if we use some other
> methods. So even if the initial estimation of the frequencies might
> not do much harm in the accurate synthesis of the sinusoids ( since we
> can adjust the freq offset with phase modulation), it might be helpful
> in qunatizing them. 
> 
> Am I missing out something here ?
> 
> ~Anup
> 
> 
> On 7/4/06, Anup K C <anupkc at gmail.com> wrote: 
>         Hi Jean-Marc
>         
>         This is what I have been trying. 
>         
>         
>