N. Aldhahir et Jm. Cioffi, FAST COMPUTATION OF CHANNEL-ESTIMATE BASED EQUALIZERS IN PACKET DATA-TRANSMISSION, IEEE transactions on signal processing, 43(11), 1995, pp. 2462-2473
Computationally efficient procedures are introduced for the real-time
calculation of finite-impulse-response (FIR) equalizers for packet-bas
ed data transmission applications, such as wireless data networks. In
such packet data applications, the FIR equalizer filters are computed
indirectly by first estimating the channel pulse response from a known
training pattern embedded in each packet and then computing the equal
izer for use in the recovery of the remaining unknown data in the pack
et We find that a minimum mean-square-error decision-feedback equalize
r (MMSE-DFE) with a finite-length constraint on its feedforward and fe
edback filters can be very efficiently computed from this pulse respon
se, We combine a recent theory of finite-spectral factorization for th
e MMSE-DFE with the theory of structured matrices to derive these effi
cient procedures for computing the equalizer settings. The introduced
method is much more computationally efficient than direct computation
by matrix inversion or the use of popular gradient or least-squares al
gorithms over the duration of the packet.