| Face recognition technique is one of the most important branches of biometrics, and it is widely applied in information security, criminal detection and import and export control. The essence of the face recognition is classification problem, the traditional classification methods often have overfitting phenomenon. Support Vector Machine (SVM) is the most popular classifier for face recognition because of outstanding learning performance and good capabilities in generalization. But when the training sample size is too big, a hot point of research is how to solve the contradictions between training speed and the training sample size. Based on evolutionary algorithm has good adaptability and concurrency, and can deal with large-scale and complicated data, the paper try to solve this problem using chaos theory and Particle Swarm Optimization.This paper mainly study face class method, which concept can be summarized as follows:1. A new circle model is proposed, the experimental results on matlab show the chaotic variables which is generated by circle model can distribute in the problem search space more evenly.2. An improved Chaos Particle Swarm Optimization (CPSO) is proposed. The particle search space of basic particle swarm algorithm is a small area, so it can not cover the entire space, and easy trapped into local optimum. Based on the ergodicity and high sensitivity to the initial value of chaotic system, combining the chaos and particle swarm algorithm to avoid that PSO getting into local optimum, and to improve the overall searching ability of the algorithm.3. A face recognition method using the above improved algorithm to train Support Vector Machine (SVM) is presented. Using the parallel computation ability of chaotic particle swarm optimization algorithm to solving the proplem of slow speed when training large-scale data. The experimental results on face database show that this method can be used to improve efficiency of training support vector machine. |
PDF Full Text Request