Abstract: Dictionary learning in sparse coding plays an important role on image recognition based on sparse representation. Considering that Gabor feature is robust to variations of expression, illumination and pose. Therefore, a face recognition algorithm via sparse representation is proposed based on Gabor feature and Support Vector Guided Dictionary Learning（GSVGDL）. At first the image Gabor features are extractes and used as the augmented Gabor feature matrix to construct the initial dictionary. The dictionary learning model combines the reconstruction error with the discrimination term and the regularization term, and formulates the discrimination term as the weighted summation of the squared distances between all pairs of coding vectors. Then a structural dictionary and linear classifier is learned simultaneously by the dictionary learning, which the learned dictionary atoms are corresponded to the class labels. GSVGDL can adaptively assign different weights to different pairs of coding vectors and enhance the discrimination of the dictionary. Experiment results show that the proposed method has good recognition accuracy and higher recognition efficiency．

Key words: sparse coding, sparse representation, Gabor feature and Support Vector Guided Dictionary Learning（GSVGDL）, Gabor feature, face recognition

摘要： 稀疏编码中的字典学习在稀疏表示的图像识别中扮演着重要的作用。由于Gabor特征对表情、光照和姿态等变化具有一定的鲁棒性，提出一种基于Gabor特征和支持向量引导字典学习（GSVGDL）的稀疏表示人脸识别算法。先提取图像的Gabor特征，然后用增广Gabor特征矩阵来构造初始字典。字典学习模型中综合了重构误差项、判别项和正则化项，判别项公式化定义为所有编码向量对平方距离的加权总和;通过字典学习同时得到字典原子与类别标签相对应的结构化字典和线性分类器。该字典学习方法能够自适应地为不同的编码向量对分配不同的权值，提高了字典的判别性能。实验结果表明该方法具有很好的识别精度和较高的识别效率。

关键词: 稀疏编码, 稀疏表示, GSVGDL字典学习, Gabor特征, 人脸识别