I have spent a lot of time tweaking different parameters. But I cannot seem to get rid of acoustic echo in my test vector. I am attaching a a sample rin and sin files.
The loudest male voice is the only voice that should be heard. Everything else is an echo.
Can somebody please help.?
They are in raw-16khz-mono format.
Really hard to define an exact delay value using the files. Easy to do when some clear(knock for ex.) sound appearsat the beginning. Nevertheless the best result I've got (uponthe files) is when delay is 0 or 59 and tail length 16.
-keeping rin-sin > 6dB.
- May be better to roze rin to have less distortions.
- Interruptions (in background noise for ex.) may give bad to AEC performance as well, forcing to bigger changes inadaptation filter coefficients.
Ivan (video/speech coding)
I mean rin magnitude amplification. I'm not sure if it will help.It may depend on the device (material). Automatic level control (ALC==AGC) usuallyused on the render side to prevent too big rin magnitude (and corresponding device vibration (led to (distorted) echo penetration via corpus). IPP provides ALC primitives.