基于CNN和SVM的人脸识别系统的设计与实现

Design and Implementation of Face Recognition System Based on CNN and SVM

针对人脸识别在实际应用中存在姿态变化、表情、遮挡等问题,研究了结合支持向量机(SVM)分类的卷积神经网络(CNN)人脸识别算法,设计并实现了人脸识别系统。系统首先使用CNN提取人脸特征向量,再将特征向量通过SVM进行分类。测试结果表明,系统在训练样本充分时面对人脸姿态变化、表情、遮挡等情况下都具有较好的性能,识别率在95%以上,能满足一般的人脸识别需求。

In order to solve the problems of attitude change,pose change,obsured in actual application of huaman face recogni⁃tion system,the face recognition algorithm based on convolutional neural network combined with support vector machine is studied,the face recognition system is designed and implemented.Firstly,the facial feature vectors are extracted by CNN,and then the fea⁃ture vectors are classified by SVM.The test results show that the system has good performance in attitude change,pose change,ob⁃sured and so on when the training samples is sufficient.The recognition rate is more than 95%,and it can meet the demand of normal use of face recognition.