Adaptive blind equalization and demodulation without channel and signal parameter extraction

Abstract

The problem of filtering a source of known statistics in noise and/or interference of unknown statistics is addressed in this paper. In a general framework, the optimum estimation (detection) of the desired signal requires perfect knowledge of the probability density functions of the interfering signal and usually leads to complex non-linear algorithms. Herein we have restricted ourselves to linear estimators and optimize performance within this subset according to a criterion derived from information-theoretic considerations. For validation purposes, the particular application of blind data demodulation has been considered. Synchronisation, carrier phase rotation and equalization are performed with the same algorithm to force a source of a known pdf at the output of the adaptive system. Performance at low SNR is derived through simulations.