This is in response to the original post, but this seems to be the
only place google would let me reply.

If x is complex then ex will still be complex so dropping the
conjugate wouldn't help.  The conjugate is needed for the complex
operations to form proper inner products.  Although the product ex*
will in general be complex, the expected value E[ex*] will still go to
zero (0+j0) - since the lms update has this form it must go to zero
otherwise the coefficients would never stop adapting.  Although if you
are consistent it really doesn't matter where you put the conjugate,
it is specified by properties of the inner product.

I don't think the complex lms is readily obtained because the complex
conjugate is not analytic.

Matt

leeshauyao@hotmail.com (Leeshauyao) wrote in message news:<fe7ef1c9.0309211656.7a02bded@posting.google.com>...
> when doing subband, you get the signal in complex value, and you do
> lms in complex value, when you reconstruct the complex e back to
> full-band e, how will you get real value back? or just take magnitude?
> 
> Lee

when doing subband, you get the signal in complex value, and you do
lms in complex value, when you reconstruct the complex e back to
full-band e, how will you get real value back? or just take magnitude?

Lee