【摘要】 单纯的支持向量机(SVM)对于人脸图像识别率并不高，并且无法快速地处理特征向量维数过高时的人脸图片，因此提出了基于粒子群算法(PSO)改进的SVM-KNN算法，该方法首先提取人脸图片的特征系数，然后利用PSO优化SVM分类器参数，用径向基(RBF)函数作为核函数，优化SVM惩罚参数C与核半径参数g,最后结合KNN算法进行分类识别。该方法在ORL人脸库中识别率达到99.40%,在Yale人脸库中识别率达到96.80%,且处理速度也优于常用的其它方法，表明该方法的有效性。更多还原

【Abstract】 Simple Support vector machine support vector machine(SVM) is not efficient for face image recognition, and it can not process the face image quickly when the feature vector dimension is too high, in this paper, an improved SVM-KNN algorithm based on particle swarm optimization(PSO) is proposed, which firstly extracts feature coefficients of face images, and then uses PSO to optimize the parameters of SVM classifier, using radial basis function(RBF) as kernel function, SVM penalty parameter C and kernel radius parameter G were optimized, and finally, KNN algorithm was used for classification and recognition.The recognition rate of this method is 99.40% in ORL and 96.80% in Yale, and the processing speed is also better than other methods, which shows the validity of this method.更多还原