Second-order statistics of the received signal can be used to equalize a co
mmunication channel without knowledge of the transmitted sequence. Blind ze
ro-forcing (ZF) and minimum mean-square error (MMSE) equalization can be ac
hieved with linear prediction-error filtering. The equivalence with the equ
alizers derived by Giannakis and Halford is shown, and adaptive predictors
that result in a lattice filtering structure are applied. The required chan
nel coefficient vector is obtained with adaptive eigen-pair tracking. Eithe
r forward or backward prediction errors can be used. The performance of the
blind equalizer is examined by simulations. The MMSE of the optimum FSE is
approached, and the algorithm exhibits robustness to channels with common
subchannel zeros.