A detection method of complexion based on SVM is proposed in this paper. The technique is an approach for locating faces in a scene image based on the detection of skin color of an unspecific human which is essentially important in development of contact free human-machine-interaction (HMI) as well as in machine visions. The simulation results show that it may not only get better rate of correct recognition of complexion than that by using traditional artificial neural network but also is better in generalization.This is because that according the criteria of structural risk minimization of support vector machine(SVM),the errors between sample-data and model-data are minimized and the upper bound of predicting error of the model is also decreased simultaneously.