A Variable Leaky Entropy-Based Whitening Algorithm for Blind Decision Feedback Equalization

This paper addresses the joint entropy maximization algorithm constrained by the variable leaky factor (JEM-VL) aiming at mitigating the feedback filter (FBF) mismatch effects characterizing the operation of the decision feedback equalizer which, at the start of adaptation, swaps positions of feedforward and feedback filters so that the latter acts as a linear all-pole whitener of the received signal. The FBF mismatch is a result of disparity between the FBF setup achieved in the blind mode by observing channel outputs and an excepted FBF setup in the tracking mode which is driven by detected data symbols. Depending on the given signal complexity and inter-symbol interference severity, the FBF filter mismatch is typically manifested by the equalizer convergence instability or even the catastrophic error propagation effects arising at the time of the equalizer structure-criterion switching from the blind to the tracking operation mode. The constraint of superfluous coefficients of the FBF filter by means of the JEM-VL algorithm eliminates the equalizer convergence instability at the time of its switching and, consequently, increases equalization successfulness. The efficiency of the JEM-VL algorithm is verified by simulations using the 64-QAM signal.