I'm using the echo canceller (I use SUBBAND/FULLADAPT with NLP/CNG/TD).
I'm having a question about the echo tail.
Let's say I want to remove an echo with a constant time delayof 150ms.
Do I have to use an echo tail of at least 150ms or a small echo tail of 16ms would be able to remove it ?
Does the FULLADAPT parameter modify the echo tailas needed ?
Genarally, in acoustic echo case the echo "tail = delay + echo replica" size, so the less delay to come to the echo canceller the smaller echo tail can be used. Basically, best case is time-aligned for the mic and speaker signals.Application specific delays like one come frombuffering, recording etc should be made as smaller as possible.
Full and lite adapt controls complexity of adaptation algorithm only.
Vyacheslav, IPP, speech coding
Thanks for your reply, I get it now.
The last parameter I was unable to find out the real meaning is the zeroCoeff (zero filter's coefficent).
Do you happen to know what it exactly does ? is it better on or off for accoustic ?
(Even the sample has a big "???" beside the parameter)
I would like to use dynamic linking, what API function call do I include in my code? The sample code uses static linking (this has given many link errors).