Abstract: This paper presents an adaptive blind support vector machine equalizer(ABSVME)based on least square support vector machine(LSSVM)with good universality. Derived from signal feature reconstruction idea,the method conducts oversampling of the output of LSSVM equalizer and designs a decorrelation cost function, with Kumar express algorithm and static iterative learning algorithm incorporated, to achieve fine online channel tracing.The effectiveness is supported by simulation in which the comparison with other nonlinear equalizers is made.The result shows the method has excellent performance in nonlinear equalization.