Reply by mossi idrissa moctar December 15, 20082008-12-15
I'm not sure that I understand what you mean.
I think that:
A :should be your input (B+noise).
B : your reference.
e: is your error use for updating your filter.
y : your output.
Maybe you could try a Wiener filter.
7 passage de la Bourgade
06410 - Biot
cel: 0033631020920
Reply by December 12, 20082008-12-12
I am implementing a LMS adptive filter denoising system. And I got a noisy signal A, and also a noise signal B. The B is corrected with noise of noisy signal A.

y(n) = conj(W)* U(n);
e(n) = d(n) - y(n);
W(n+1) = W(n)+mu*U(n)*conj(e(n)) ;

1. Which signal should act as main input signal U,and which should act as the desired signal D? noisy or noise? why? can I exchange their roles? Please tell the principle.

2. Which is the result,the denoised speech signal? e or y? Why?