ABSTRACT
Through the analysis of the basic theory of human face recognition and the analysis of the Kernel Independent Component Analysis (KICA) method, the directed acyclic graph method is used to solve the multi-class problem in face recognition, in this paper. A new mixture kernel function is constructed on the basis of the polynomial kernel function and radial basis kernel function, and introduced into the classification decision. Experiments were carried out on the ORL face database. The classification recognition rate under different nuclear parameters is analyzed. The optimal kernel parameter value of the mixture kernel function is obtained and combined with the support vector machine method, which is used for classification and face recognition.
