自适应Gabor特征的协同表示人脸识别算法

为了解决人脸识别应用中针对人脸姿态的变化,光照等外部环境变化导致识别率不高,且稀疏表示应用于人脸识别收敛速度慢的情况,提出了一种基于多分量的Gabor特征提取和自适应权重选择的协同表示人脸识别算法(GAW-CRC).特征提取阶段,将Gabor变换的所有特征分量中鉴别能力较差的分量淘汰,剩余分量构建特征字典,分别协同表示对应测试样本的特征分量,将所有剩余分量的识别结果,按照自适应的权重函数加权融合得出最终分类结果.实验证明:算法应用于AR,FERET与Extended Yale B人脸库中,当对应的样本存在人脸角度变化,表情变化和光照条件变化等情况时,能够得到更高的识别率.

Adaptive Gabor feature extraction and collaboration representation face recognition algorithm

In order to solve the problem that in face recognition applications posture change of human face and external environment change such as illumination results in low recognition rate,and slow convergence speed and sparse representation used in face recognition,a face recognition algorithm based on multi-component Gabor feature extraction and adaptive weight selection collaborative representation(GAW-CRC)is proposed.In feature extraction stage,the method eliminates component that has the weakest identify ability of all the Gabor transform feature components. The remaining components build feature dictionary,combining the feature component of corresponding test component,will have a final result by weighted merging together recognition results of all the remaining components according to the order of self-adaptive weighting function. Experiments show that this algorithm is possible to obtain a higher recognition rate in AR,FERET and Extended Yale B face database when the corresponding sample have changes in face angle,expression lighting conditions and other circumstances.