Face recognition using support vector machines with the feature set extracted by genetic algorithms

Abstract

Face recognition problem is challenging because face images can vary considerably in terms of facial expressions, 3D orientation, lighting conditions, hair styles, and so on. This paper proposes a method of face recognition by using support vector machines with the feature set extracted by genetic algorithms. By selecting the feature set that has superior performance in recognizing faces, the use of unnecessary information of the faces can be avoided and the memory requirement can be decreased significantly. Also, by using a tuning data set in the computation of the evaluation function, the feature set which is less dependent on illumination and expression can be selected. The experimental results show that the proposed method can provide superior performance than the previous method in terms of accuracy and memory requirement. © Springer-Verlag 2001.