Abstract: Support Vector Machine（SVM） is an effective learning tool for small size samples.The SVM is considered as a good classifier with high generalization performance with no need to add a priori knowledge.However，the performance of SVM is influenced greatly by the scale parameter σ² and the penalty parameter C，therefore，Genetic Algorithm（GA） to improve SVM in selecting these parameters based on face recognition is presented in this paper.Firstly，features from human face images are extracted by combining the 2-D wavelet decomposition technique with the grayscale integral projection technique，and then，the optimized SVM to recognize is applied.The experimental results show that the proposed approach is efficient in face recognition.

摘要： 支持向量机是一种基于小样本学习的有效工具，作为分类器被认为具有很高的推广性能，无需先验知识。但是参数的选取与支持向量机的识别性能是相关的，核函数参数σ²和惩罚因子C对支持向量机识别性能会产生很大的影响。针对支持向量机在人脸识别问题中的应用，提出了一种基于遗传算法（GA）的参数选择优化方法。利用笔者曾提出的基于小波分解和积分投影的人脸特征提取算法对人脸图像进行特征参数提取，然后利用优化的支持向量机进行识别。实验结果表明，该方法是有效的。