Fractionally Spaced Bayesian Equalizer

In digital communication system, conventional fractionally spaced equalizer (FSE) is based on linear transversal equalizer (LTE) where the noise enhancement influences the equalization result. Since equalization can also be viewed as a classification problem, the Bayesian equalizer which is the optimal symbol-decision equalizer can achieve better performance than LTE. A novel fractionally spaced Bayesian equalizer (FSBE) is proposed in this paper. Corresponding to sampling received signal at N times symbol rate, N parallel Bayesian equalizer can be designed. The decision variable in FSBE is defined as the product of the Bayesian decision variables in these N Bayesian equalizers. The proposed FSBE presents better BER characteristics than Bayesian equalizer. The validity of the proposed FSBE is demonstrated by the simulation.