Welcome! Log In Create A New Profile

Advanced

Feature Request: Improved Echo-cancelation Algorithm

Posted by parham 
Feature Request: Improved Echo-cancelation Algorithm
March 06, 2012 08:18AM
Hi there,

I was wondering if there's a way to improve the echo cancelation algorithm used in TT. I have used the same computers to have a voice-conference on TT and, say, Skype, and I get no echo on Skype. Is there a way to somehow replicate something like that algorithm before Microsoft patents it? :-)

Thanks.
Re: Feature Request: Improved Echo-cancelation Algorithm
March 06, 2012 05:23PM
Hi

When you enable echo cancellation also enable Audio Config in the Create Channel dialog. This help to improve echo cancellation. It, however, doesn't work as well as in Skype.

-- Bjoern
Re: Feature Request: Improved Echo-cancelation Algorithm
March 07, 2012 06:21PM
I think that there's no need to improve this algorithm.
Skype can greatly remove echo, because it (when echo occurs) limits the duplex between one and second side.
Silk - codec built into skype splits signal to many bands, and tells the algorithm, which bands to filter.
You can also consider that Echo cancellation in teamtalk works better with noise reduction turned on, because EC algorithm can have more clear comparison material.
Also Noise reduction without EC can remove or attenuate some echo artifacts before applying cancellation.
Re: Feature Request: Improved Echo-cancelation Algorithm
March 11, 2012 02:48PM
We have the same problem using handsfree, we tried to change its operation imitating the behavior of skype, but without great success.
We have introduced a little delays and a lowering of the signal received on the client but I believe it can be much improved if this control can be managed by the server before sending the stream.
Unfortunately the echo cancellation seems not work.
We also used the handsfree Polycom systems with echo cancellation but without any appreciable result.
This is a great point which should be surely improved.
My opinion is that with skype handsfree operation is far superior and there are no Larsen effects.
More effort in this direction to solve this problem would be desirable, make this library much more usable.
We have the need to have in some cases the use of the hands-free on both sides and this creates significant problems for its use.



Edited 2 time(s). Last edit at 03/11/2012 02:51PM by indiagolf.
Re: Feature Request: Improved Echo-cancelation Algorithm
March 12, 2012 07:30AM
There is room for improvement in echo cancellation. The simplest solution is to use whenever possible headsets. Alternatively, you can try using add-on software to reduce the echo. One add-on that I am familiar with (SoliCall Pro) can remove both near-end echo and far-end echo (i.e. echo generated at the far end).
Re: Feature Request: Improved Echo-cancelation Algorithm
March 12, 2012 08:12AM
I have looked at SoliCall Pro's website. However, I don't understand how I may use it in conjunction with TeamTalk. Should I get the TeamTalk source and edit it to use the SDK?

Thanks!
Re: Feature Request: Improved Echo-cancelation Algorithm
March 12, 2012 02:42PM
I was reffering to a more easy way of improving audio on a single PC. Once installed SoliCall Pro will act like a regular audio device on your PC. You can use this new audio device when making calls. I guess there is also the option of integrating the SDK but it looks to me that the SDK is more suitable for vendors.
Re: Feature Request: Improved Echo-cancelation Algorithm
June 05, 2012 04:35PM
Hi

I've added support for Windows Vista/7 latest sound system, so you now should have improved echo cancellation and latency. You can try it here:
[bearware.dk]

-- Bjoern
Re: Feature Request: Improved Echo-cancelation Algorithm
June 10, 2012 03:54PM
Hi Björn.
Thanks for adding WASapi. now teamtalk has fine latency, and it's ok.
I found problem with that thing.
On most win 7's (we used celt VBR, 44.1 K stereo, 192 kbps channel) WAS api transmits signal, it's about minute or two, than loops last queued buffer until you will rejoin.
Length of buffer is about 20 - 50 MS (Note F) and there's no solution than leaving and joining again the same channel.
It occurs when you use voice activation (act level set to 0, but even, when PTT enabled, so i think, that it doesn't mind.
In celt mono (where i tested 3D positioning, which isn't available but i know it because of API specification) also there are loops.
Also you can add device latency slider when audio session selected, because some cards can transmit and receive with for example 30 ms allocated on device, some with 10 when using WASapi.for example on some VIA HD audio deck, WASapi drops when receiving, you can hear clicks, but one of my realtek cards on netbook, can go much lower, than it's set in teamtalk.

Thanks for adding it, and i hope, that you will resolve this problem, because now, it's barely usable. (you must leave/join every 2, 3, max 5 minutes).
#remark: Changing transmit interwal doesn't solve this issue.

Kind Regards: Patryk
Re: Feature Request: Improved Echo-cancelation Algorithm
June 10, 2012 04:50PM
Hi Patryk

Do you use duplex mode?

-- Bjoern
Re: Feature Request: Improved Echo-cancelation Algorithm
June 19, 2012 07:42PM
As i see, the Duplex mode isn't available.
when using windows audio session.
In teamtalk 4.4 B1 problem also occurs.
Also as i said earlier is it possible to add device latency allocation slider?
Various cards in WAS need various latency.
On some it can be less, than that set currently in teamtalk.

Regards - Patryk.
Re: Feature Request: Improved Echo-cancelation Algorithm
June 19, 2012 10:10PM
Hi

For duplex mode to be active both input and output device must use the same sample rate. You can change this in your sound device settings in Control Panel.

In TT it's the codec which determines the latency. If you increase the latency, does the playback become more stable?

-- Bjoern
Re: Feature Request: Improved Echo-cancelation Algorithm
June 27, 2012 10:29PM
Hi Patrick

Could you please try the following client and see if it gives more stable playback:
[bearware.dk]

Do not enable duplex mode, since this is only a test version.

-- Bjoern
Re: Feature Request: Improved Echo-cancelation Algorithm
June 29, 2012 05:35PM
Server Name: Flea markets KRAMATORSK IP ADRESS 77.121.251.136 UTP & TCP: 10333 NAME ON SERVERVE 1 - [rtz-2.ucoz.org]
Sorry, only registered users may post in this forum.

Click here to login