[Speex-dev] Sound card problem in acoustic echo cancellation

Anton A. Shpakovsky saa at tomsksoft.com
Wed Jun 9 18:03:53 PDT 2010



Just want to share some experience. 

I¡¯ve tried several commercial AEC solutions ¨C negotiations, signing NDA ¨C but still no better results than speex get.


Now I¡¯m working with GIPS SDK ¨C I heard that Skype do not use GIPS VoiceEngine anymore, but yes ¨C WLM, Yahoo and AOL messengers are using GIPS VE.

And it is not only echo suppression. Their SDK is very powerful and I hope soon I will get some positive results.

As for now ¨C SPEEX is the best.




Anton A. Shpakovsky


From: speex-dev-bounces at xiph.org [mailto:speex-dev-bounces at xiph.org] On Behalf Of limaoquan2000
Sent: Thursday, June 10, 2010 2:14 AM
To: speex-dev at xiph.org
Subject: Re: [Speex-dev] Sound card problem in acoustic echo cancellation


Then why ONE sound card have different capture and playback rate?
It must be ONE single physical clock generator which is used by both ADC and DAC
in the sound card, isn't it?
If you are a hardware engineer. Will you design two different physical clock for
ADC and DAC seperately?
What on earth causes this problem? Who knows its intrinsic real reason?

Isn't there any other solutions?
For example, I have also tested Microsoft MSN and Skype. It seems that AEC
function of those softwares are all no problem. Their AEC are designed by GIPS.
I guess that must be some solutions.
The solution of GIPS is only echo suppression?

Em 02-06-2010 04:41, ce09 at polyu.edu.hk escreveu:
> Hi All,
> I am a research associate in the Hong Kong Polytechnic University. One of
> my research interests is acoustic echo cancellation. Now I meet a big problem.
> When I was testing my own AEC module, I found that it was almost perfect in
> few computers but much worse in other computers. Then I tried AEC module in
> speex, it was almost the same result.
> Then I found the reason is that sound cards of most computers have different
> sampling and playing rate, so it is almost impossible for the adaptive filter
> to converge and reach best result. For example, long time average sampling and
> playing rate of a sound card is 7999.864Hz and 7980.129Hz respectively. Sound
> cards of the computers which shown perfect AEC result are the only sound cards
> which keep the same sampling and playing rate.
> Then, what can I do to solve the problem? Could you please give me any advice?
> BTW: I am working on Windows XP platform and I have tried DirectSound.
> Your Sincerely
> Li Maoquan


ÍøÒ×ΪÖÐСÆóÒµÃâ·ÑÌṩÆóÒµÓÊÏ䣨×ÔÖ÷ÓòÃû£© <http://ym.163.com/?from=od3>  

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20100610/6b3508ff/attachment.htm 

More information about the Speex-dev mailing list