I think this should be of interest even to people not interested in speech compression. I have just released an acoustic echo canceller (AEC), as part of Speex (http://www.speex.org/) 1.1.9. While the AEC has been there for a while, this is the first version that actually works. For those not familiar with AEC, the idea is that when doing full duplex, it removes the speaker signal from what is captured by the microphone. I should also mention that this work was sponsored by Tipic Inc (http://www.tipic.com/).
Documentation on how to use the AEC is available in the manual.pdf file (in the source distribution).
In the source there is a function:speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise).In these parameters, the "echo_buf" means the speaker signal from what is captured by the microphone.My question is how could I get the echo and record it to a file.
In the source there is a function:speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise).In these parameters, the "echo_buf" means the speaker signal from what is captured by the microphone.My question is how could I get the echo and record it to a file.
[a href="index.php?act=findpost&pid=367625"][{POST_SNAPBACK}][/a]
The only way for echo cancellation to work correctly is when both ref_buf and echo_buf are obtained simultaneously and go through the same soundcard. Otherwise, sampling rate drift will make that impossible.
In the source there is a function:speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise).In these parameters, the "echo_buf" means the speaker signal from what is captured by the microphone.My question is how could I get the echo and record it to a file.
[a href="index.php?act=findpost&pid=367625"][{POST_SNAPBACK}][/a]
The only way for echo cancellation to work correctly is when both ref_buf and echo_buf are obtained simultaneously and go through the same soundcard. Otherwise, sampling rate drift will make that impossible.
Hi,
I would like to know that in case of playing audio and capturing from the mic at the "same time"... isn't there time associated with thread switching and also room's reverberation rate (acoustic). Then how can it be determined that both the frames are simultaneous??? Or is that time is taken into considerations???
I am using speex in my voip based product....kindly answer my question Jean
In the source there is a function:speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise).In these parameters, the "echo_buf" means the speaker signal from what is captured by the microphone.My question is how could I get the echo and record it to a file.
[a href="index.php?act=findpost&pid=367625"][{POST_SNAPBACK}][/a]
The only way for echo cancellation to work correctly is when both ref_buf and echo_buf are obtained simultaneously and go through the same soundcard. Otherwise, sampling rate drift will make that impossible.
Hello! You just told that "both ref_buf and echo_buf are obtained simultaneously and go through the same soundcard." But in full duplex,how do u abtain 2 signal from a "same" soundcard? It is impossible! Two signal,one is refence signal ,which is transimitted from far side by network, captured by "that" far Mic , sampled and encoded by "that" far side codec,decoded by "this" near side codec. Another signal, it is echo signal,captured by" this" near side Mic and sampled by this codec. How does this 2 signal go through the same soundcard?
Can u kindly explian it?
now I am studying a PBFDAF AEC with full duplex mode in PC/PDA ,encounting some difficulty. Maybe "sampling rate drift "can open my heart . Later,I would study speex AEC and impelment it.
In the source there is a function:speex_echo_cancel(st, ref_buf, echo_buf, e_buf, noise).In these parameters, the "echo_buf" means the speaker signal from what is captured by the microphone.My question is how could I get the echo and record it to a file.
[a href="index.php?act=findpost&pid=367625"][{POST_SNAPBACK}][/a]
The only way for echo cancellation to work correctly is when both ref_buf and echo_buf are obtained simultaneously and go through the same soundcard. Otherwise, sampling rate drift will make that impossible.
Hi,
I would like to know that in case of playing audio and capturing from the mic at the "same time"... isn't there time associated with thread switching and also room's reverberation rate (acoustic). Then how can it be determined that both the frames are simultaneous??? Or is that time is taken into considerations???
I am using speex in my voip based product....kindly answer my question Jean
AEC do not demand two signal absolutely "simultaneous"! "absolutely simultaneous" is impossible because of thread switching. AEC can work well, although two signals misalign for some degree sample data (shorter than filter length) . But ref signal must go ahead echo signal.
where is the source code????
where is the source code????
?? Did you read the thread? http://www.speex.org/downloads/ (http://www.speex.org/downloads/) , source file mfd.c + the rest
Hi,
I am using speex aec algorithm, I can't able understand theory behind the algorithm. So please, anyone suggest me where can I find the detailed description docs or
can any one give description about the algorithm?
Thanks
Ramesh Nagapuri