Soft decision approaches for blind adaptive decision feedback equalizers

We consider joint blind optimization schemes for decision feedback equalizer (DFE) coefficients and decision device, and propose adaptive solutions based on a soft decision device which combines a blind infinite impulse response (IIR) filtering adaptation and the conventional decision directed (DD) DFE adaptation. The benefits of our approach are mainly three-fold: (1) the initialization of DFE taps does not rely on a training sequence; (2) the transition between the (re)initialization period and the tracking period is continuous and consequently the equalizer exhibits robust performance under sudden time variation in the multipath channel; (3) the optimized soft decision device mitigates error propagation of the DFE and further reduces the mean square error (MSE) of the DFE.