基于混合Gauss模型的鲁棒性人脸识别算法

针对人脸图像受表情、光照、角度变化等因素影响,传统算法难以获得较理想的人脸识别结果问题,提出一种基于混合Gauss模型的鲁棒人脸识别算法.先将每副图像划分成子块,提取其方向梯度直方图特征,并加入子块相应的空间位置信息产生人脸图像的局部特征向量;再采用全部图像的局部特征向量训练混合Gauss模型生成人脸特征向量;最后采用最小二乘支持向量机建立人脸识别分类器,实现人脸匹配与识别.采用ORL,Yale和CIGIT人脸库进行仿真对比测试,仿真结果表明,该算法的人脸识别率高于其他人脸识别算法,对光照、角度、表情等有较强的鲁棒性,且可以获得更快的人脸识别速度.

A Robustness Face Recognition Algorithm Based on Gaussian Mixture Model

It is difficult to use the traditional algorithm to achieve the ideal result of face recognition under the influences of light, angle and other negative factors and this paper presents a robust face recognition algorithm based on Gaussian mixture model. First of all, ach of the sub image is divided into sub blocks, and the orientation gradient histogram feature is extracted, and corresponding spatial location information of sub block is used to generate local feature vectors of face image. Secondly local feature vectors of all the images are used to train Gaussian mixture model to generate feature vectors. Finally, least squares support vector machine is used to build classifiers for face recognition and match and recognize face. The simulation test was carried out with ORL, Yale, and CIGIT face database, the simulation results show that the face recognition rate of the proposed algorithm is far higher than those of the contrast face recognition algorithms, and it has stronger robustness for light, angle and expression, and can obtain the faster face recognition speed, and has higher practical values.