Blind equalization of short burst signals based on twin support vector regressor and data-reusing method

Abstract

In this paper, blind equalization of short burst signals is formulated with the twin support vector regressor (TSVR) framework. The proposed algorithm combine the conventional cost function of TSVR with classical error function applied to blind equalization: the Godard's error function that describes the relationship between the input signals and the desired output signals of a blind equalizer is contained in the penalty terms of TSVR, and the iterative re-weighted least square (IRWLS) algorithm is used for twin support vector regressor to achieve fast convergence. In addition, it utilizes the data-reusing method for small amounts of data samples to reach stable convergence. Simulation experiments for constant modulus signals are done to prove the feasibility and validity of the proposed algorithm.