Turbo block Bayesian decision feedback equalizer

Turbo equalizer (TEQ) is well known for its tremendous performance gain over a severe intersymbol interference (ISI) channel. However, the original MAPSE-based (maximum a posteriori symbol estimation) equalization algorithm is still not practical for implementation under long channel memory and high modulation level scenarios. In this paper, we consider using the much simpler decision-feedback equalizer (DFE) structure designed by a Bayesian MMSE (BMMSE) criterion to tackle the above computational problem. In so doing, the DFE makes use of the feedback extrinsic likelihoods from the decoder as the estimate of the a priori probabilities of transmission symbols, and BMMSE design formulae are derived for symbol-updating of the feedforward and feedback filter coefficients. To further reduce the computational load, a turbo block Bayesian decision-feedback equalizer (TBB-DFE) is then proposed which updates the DFE coefficient only once a block. The simulation results show that for a severe 5-tape ISI and 8-PSK case, while the conventional nonturbo DFE receiver completely fails, and the MAPSE has too many trellis states, our TBB-DFE receiver can attain a BER of 10^-3 at an E_b/N₀ of 19 dB after 6 iterations. Thus the TBB-DFE can serve as a low-complexity TEQ approach under high-level modulation and long channel memory situation.