A stochastic whitening algorithm with time variable neuron slope for blind DFE

The self-optimized blind decision feedback equalizer (DFE), which independently optimizes feed-forward and feedback filters, deals with the feedback filter mismatch. To mitigate the filter mismatch effects, we have proposed a stochastic entropy gradient algorithm with a time variable neuron slope for all-pole filter which acts as a front-end amplitude equalizer in the blind operation mode. The algorithm performances are verified by simulations for 16- and 64-QAM (quadrature amplitude modulated) signals.