Ipp 5.0 Echo canceller

[duane]

Has anyone had any success integrating this? How do you handle the fact that typically the microphone is separate from the speaker, whereas the echo canceller API requires both in the same function call? Are people typically holding one frame of speaker output around, and using it with the EC function call when the microphone data is dispatched from the operating system?

[Intel_C_Intel]

Generally, the mic input is treated as the mixed signal and the (filtered) speaker out put as the echo. To use AEC you should control in your application or at least get as the mic input as the speaker output

[Antonello_Rizzi]

I have implemented a SIP Phone with ixp425 pocessor,
using DSR 2.6.2 and AccessLibrary 1.4 and i have acoustic
echo problem.

I have captured voice streams in a VoIP conversation with ethereal
obtain two wav file: reversed.wav (speakerphone signal) and forward.wav
(microphone signal).

I have tried to use the sc_ec example (IPP 5.0 linux Speech Coding)
to remore echo. I used sc_ec as following:

[voip-phone@voip.test]# ./usc_ec -ah -a fb reversed.wav forward.wav forward_filter.wav
Staticly loaded echo canceller: EC_INT
The Intel echo canceller conformant to ITU G167 and G168,
The Intel IPPSC library used: 5.0.11 Build 72, name ippscs2.a
The Intel IPPSP library used: 5.0.11 Build 100, name ippss2.a
Input rin file: reversed.wav
Input sin file: forward.wav
Output sout file: forward_filter.wav
0.07 MHz per channel
Done 165623 samples of 8000 Hz PCM wave file (20.7029 sec)
Completed !

The output file forward_filter.wav seems to be identical to forward.wav. Why?

Where I can find audio files for testing AEC usc_ec demo application?

[Antonello_Rizzi]

I have more investigated my problem.
The usc_ec application work fine, and remove echo
with delay of 180 ms (my situation).
The problem I have is that the speakerphone samples
is not synchronized with microphone samples.

Have the IPP library a pimitive for samples Synchronization ?
My goal is insert a "sample Synchronization
Module" before AEC Module.

[Ivan_Ryzhachkin__Int]

I can suggesttrying ippsResample()for mike and loudspeaker sampling frequency alignment.