A decision-directed Bayesian equalizer

This paper investigates a new decision-directed Bayesian equalizer based on the blind Bayesian equalizer introduced by Iltis, Shynk, and Giridhar (1991, 1992). The blind equalizer of Iltis suffers from high complexity making it impractical for software implementation of QAM signals beyond 4 states and with bit rates greater than 16 kbps. The complexity of the decision-directed Bayesian equalizer is compared to the blind version. It is shown that even though the decision directed implementation is less complex, implementation of it in software still seems limited to BPSK and QPSK with bits rates less than 30 kbps. It is also demonstrated through simulation results that even for initially poor channel estimates, the decision-directed equalizer still converges to a close approximation to the actual channel impulse response.