Face recognition system can be improved in two ways:increasing recognition rate and increasing recognition speed.Recognition rate depends on the feature selection and the space dividing.Chin is a stable feature in front-view face image.Using chin as a new feature can improve the recognition rate.Further,huge database of human faces is organized in different categories with the chin features.So the consequent recognition module will work on a group of smaller database,and the correct rate and the speed of recognition will be improved.This paper introduces that how to get the chin contour,how to cluster the chin contours and how to use the cluster result in face recognition system.The experimental results show this approach improves the performance of face recognition system with huge database greatly.