Need help understanding use of dsp adptive lms filter. Topic is solved

[shaileshwankhede]

Hello,

I am trying to implement echo cancellation on xCORE Array Microphone Kit. I decided to use adaptive filter but getting confused on what value to pass for reference_sample parameter in api dsp_adaptive_lms?? I can understand we need to find transfer function of environment so that we can mimic echoed signal and thus can be cancelled out. To do this we need to iterate process over some samples and for every samples need to call dsp_adaptive_lms iteratively based on number of taps for filter. First parameter will be mic input sample value but what will be second parameter (reference_sample) in real environment as I don't know what is correct desired signal. Do I need to feed test signal like sine wave first so that I know desired signal or reference signal? And how to choose value of mu?

Thanks,
Shailesh

[johned]

Hi Shailesh,

Using an adaptive filter is quite a complex operation so first off I would refer you to a number of references :
https://www.google.co.uk/webhp?sourceid ... troduction

Specifically, to answer your questions. The filter will adapt to the reference sample. So for example if you are playing audio out of loudspeakers then this is your reference. Depending on your specific application you should not need to train it with a known signal. If your application requires this then a broadband signal is much better than a narrow band sinusoid.

Choosing the value of mu is the subject of many research papers and there is no simple answer. I would start with the values in the lib_dsp app note and see how you get on.

Best regards,
John

[shaileshwankhede]

Thanks John for insight!
One clarification needed here...Does mic array kit with default usb audio application has echo cancellation already taken care?
I get this doubt because when we connect external audio amplifier to Audio Out of mic array kit and using kit as USB audio in & out device for call application, it works seamlessly.
Even though there is caller sound on speaker, it doesn't fed back through mic again to caller. Is it taken care in hardware or effect is not so prominent that its not observed?

Thanks,
Shailesh

[johned]

Hi Shailesh,
It is interesting to hear that there is no echo but the standard USB software does not include ECAN. ECAN is included in the smart mic + firmware.
Best regards,
John