Hi all,
I'm doing a project to do 'real time active noise cancellation'
on blackfin 533.
My second step towards the project (the first step is verification of algorithm
in matlab) was to send two correlated low frequency audio signals (which i
created from matlab) into blacfin 533, and tried to cancel the signal using LMS
algorithm.
As the codec for blackfin 533 sampled the audio signal at 48kHz, i found it hard
to get 'converging mean-squared error' for the noise cancellation
system. Is the 'too fast' sampling rate make the two signals look
uncorrelated in the blackfin 533 processor's view?
what is the suitable filter lenght (i used FIR filter in the algorithm) in this
case? i tried 32,64,128 and 256 but all seems to give unsatisfactory results.
