Forums

OFDM Channel Estimation Wiener Filtering

Started by sunshine.gladys March 9, 2007
Hi all,

I am rather new to this and would appreciate if someone would kindly
enlighten me on this:

My pilots symbols are inserted into a set of subcarriers (comb type).

I've seen the MMSE estimate of the FFT of the channel coefficients given
as: H_mmse = R_HHls(R_HlsHls + sigma^2(XX^h)^-1)^-1 Hls

where Hls is the channel estimates from the pilot symbols using the least
square method. H is the FFT of the channel coefficients and R_HHls denotes
the cross correlation matrix of H and Hls. sigma^2 is the noise variance. h
denotes hermittian.

I would like to know as wiener filter minimizes the MSE, if I had used:
R_HlsHls(R_HlsHls + sigma^2(XX^h)^-1)^-1 Hls instead, meaning I only have
the channel estimates at the pilot symbols, can i use any low pass filter
to interpolate (using resample function in matlab)? Will it result in
MMSE? Or can any one kindly share with me how to construct the wiener
filter? Why i did this is because the previous expression for H_mmse did
not give me good performance. In fact interpolating using a separate low
pass filter gives me better performance.

Any insights/comments would be very much appreciated.

Thank you
Gladys