FAST COMPUTATION OF CHANNEL-ESTIMATE BASED EQUALIZERS IN PACKET DATA-TRANSMISSION

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.