Abstract: SVM has slow training speed in the condition of a great deal of samples,and support vector Pre-extracting can solve this problem.Adaboost algorithm emphasizes error samples,and error samples are boundary samples which can construct support vector.So the support vector pre-extracting method based on error samples AdaBoost algorithm is proposed.Through AdaBoost upgrading process the algorithm gets error smples a big weight,and the support vector pre-extracting carries out,in the end the simulation experiments validate the algorithm is effective.