中文摘要：

针对人脸识别系统无法识别人脸图像是否来自真人的问题,首先改进了傅里叶频谱（Fourier spectrum,FS）特征的人脸活体检测方法,并验证了局部二值模式（local binary pattern,LBP）特征和灰度共生矩阵（gray-level co-occurrence matrix,GLCM）特征人脸活体检测性能。在此基础上,提出了融合LBP特征的FS-LBP特征人脸活体检测算法。实验结果表明,提出的FS-LBP特征在多数据库的混合数据的准确率高达83.17%,更优于多尺度局部二值模式（multiscale local binary pattern,MSLBP）特征。

英文摘要：

For the face recognition system,it is difficult to recognize whether the human face image comes from a real person. Firstly,FS feature method was improved,and LBP feature method and GLCM feature method for face antispoofing were discussed respectively. Furthermore,two feature fusion techniques were presented based on FS feature and LBP feature,that is FS-LBP method. Experimental results showed that the accuracy of FS-LBP proposed in this paper is 83. 17% higher than that of MSLBP.