Support Vector Machine-Based Blind Equalization for High-Order QAM With Short Data Length

In this paper, the problem of blind equalization of high-order quadrature amplitude modulation (QAM) signals is tackled by using a batch equalizer based on support vector regression (SVR). Anew set of error functions weighted by neighborhood symbol decisions and augmented by generalized power factors p and q, are proposed to be used as the penalty terms in SVR, and the optimal values of p and q are determined. In addition, we propose a method to remove the high online computational complexity incurred by the inclusion of neighborhood terms in the new error function. Simulation results show that with about the same complexity, the optimized SVR-NA-SBD-(p, q) attain much lower residual inter-symbol-interference and higher probability of convergence than the best known SVR-MMA, and it needs only about 1400 symbols to achieve a BER of 10-4 for 256QAM in a multipath channel. In contrast, the conventional SVR-MMA needs more than 4000 symbols to achieve such BER.