Blind equalization algorithms based on high order statistics (HOS) are suitable to single input single output (SISO) systems

but HOS algorithms need large amount of data and cannot meet the time-varying requirements in high-speed signal transmission systems and have not enough ability to detect blindly high-order quadrature amplitude modulation (QAM) signals.A novel cost function is constructed using support vector regression by structural risk minimization principle whose empiric risk is composed by the cost functions of constant modulus algorithm (CMA) and constellation match error (CME)

and the epsilon-insensitive loss function is adopted

all of which transform the blind equalization problems of high-order QAM signals into solving an unconstrained optimization problem.Finally

64-QAM signal is used to simulate and analyze the performance of the novel algorithm and the results show that the burden of algorithm and the data requirements are superior to those existing HOS algorithms.