MMSE is a widely used linear estimator for i.i.d signal. If the realization of signals at different time instances are not independent, can MMSE be used or not? Does it have to do with the whitening processing? Or does some other linear estimator if valid for the non-independent identical distributed source?
Can you specify more precisely what would you like to do? Maybe I can give you some advice. Saso
Hi, if i understand your question, maybe you can use the generalized estimator for the linear model derived from the CRLB theorem. This estimator is generalized for colored noise, using a Cholesky decomposition of the noise covariance.
To Saso,
I am considering this model: Y=hX+N, where X has memory (nonwhite spectrum), N is AWGN. Can I also directly use the LMMSE or MMSE formula to calculate the estimation of X? If yes, there is any discussion on the block size of the data Y used for the estimation. Meaning, there should be any requirement on the value of L, for Y=[y_1, y_2,...y_L], when doing the estimation. Thanks.
To Rodgrigo,
Appreciate your help. I am wondering whether that method is applicable for colored signal, not noise. Thanks,
-pengda
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing