Interesting questions...There are two questions here and let me see if I can
address them.
The first question is the issue of the noise at the output of the matched
filter.
Note that you are not using EVERY output of the matched filter. You are
using only the output of the matched filter sampled every T (where T is the
symbol period). The noise samples every T are uncorrelated Gaussian random
variables (and hence independent). Proakis shows this in the section on
"Correlation Demodulator" right before he shows that the matched filter is
equivalent.
The second question is a more general question of whether the Viterbi
Algorithm can be used with correlated noise.
In this regard, let me point out an excellent paper that delves into the
Viterbi Algorithm in more depth than Proakis does. It is "The Viterbi
Algorithm" by G. David Forney, Jr, in the Proceedings of the IEEE, vol 61,
No 3, March 1973. As he says, in the most general sense, the "VA may be
viewed as a solution to the MAP estimation of the state sequence of a
finite-state discrete time Markov process observed in memoryless noise"
hope this helps...
Nandan
On 4/17/07, l...@hotmail.com wrote:
>
> Hello everyone.
> Could I have some discussion, if you are interested, please?
>
> In communication books such as Proakis, there are formulas for MLSD or
> Viterbi equalization.
>
> On the other hand, if we have a matched filter at the receiver, we'll
have
> a correlated noise sequence. Without whitening the correlated noise, can we
> still use a Viterbi equalizer?
>
> In other words, can the Viterbi equalizer process the correlated noise
> sequence?
>
> In my understanding, MLSD can be derived from a joint pdf, and this joint
> pdf can be separated as a multiplicaiton of uncorrelated Gaussian
noise'
> pdf.
>
> If we have correlated noise, we may not be able process it by Viterbi
> equalizer, I think.
>
> I'd highly appreciate if anyone could give me advice.
>
> Thank you so much in advance.
>