Blind Equalization Based On Support Vector Machine

The technique of equalization is the effective method for removing the ISI from communication systems. Blind equalization need not training sequence, thus many scholars and experts begin to study on it.Japanese Sato put forward Blind equalization algorithm at first in 1975. At present, study on Blind equalization focus on finding a method to search the global minimum point or choosing equalizer's structure, on the other word, it is directness or indirect using the high-order statistics under the same theory. Typically, such as Bussgang blind equalization, based on high-order statistics blind equalization and nonlinear blind equalization, and so on.Every algorithm of blind equalization has its applicability and according to some suppose condition, and every algorithm for blind equalization has its advantage and disadvantage. On the whole, performance improving of one algorithm in one aspect commonly based on the other lose. In the communication systems, how to chose the method to use need compromise for all of the algorithm.Learning by machine is a new domain recently, especially after the Statis study theory offer by Vapnic. VC-Dimension and SRM(Structural Risk Minimization) has put forward the base of statistics under the short data sample condition, and SVM(Support Vector Machine) is the typically method for solve such questions.In this paper, blind equalization achievement has been summarized, and SVM has been used to implement blind equalization, classify and regression method for SVM in the simulation exhibit superiority over other blind equalization methods.