基于低秩子空间投影和Gabor特征的稀疏表示人脸识别算法

目前的人脸识别算法常常忽视训练过程中噪声的影响,特别是在训练数据和待测数据都受到噪声污染的情况下,识别性能会明显下降。针对含有光照变化、伪装、遮挡及表情变化等较大噪声的人脸识别问题,提出了一种基于低秩子空间投影和Gabor特征的稀疏表示人脸识别算法。该算法首先通过低秩矩阵恢复算法得到训练样本的潜在低秩结构和稀疏误差结构;然后利用主成分分析法找到低秩结构的Gabor特征所在低秩子空间的变换矩阵;再通过变换矩阵将所有样本的Gabor特征向量投影到低秩子空间上,在该低秩子空间上使用稀疏表示分类算法进行最终的分类识别。在Extend Yale B和AR数据库上的实验表明,新算法具有较高的识别率和较强的抗干扰能力。

A face recognition algorithm based on low rank subspace projection and Gabor feature via sparse representation

At present, face recognition algorithms often ignore the influence of noise in the training process, especially when the training data and the testing data are corrupted, so the recognition performance may be decreased significantly. To solve the problem of face recognition with illumination variation, occlusion, camouflage and expression variation, we propose a new face recognition algorithm based on low rank subspace projection and Gabor feature via sparse representation. Firstly, we obtain the potential low rank structure and sparse error structure of training samples by the low rank matrix recovery algorithm, and the transformation matrix of the Gabor feature of the low rank structure by using the principal component analysis. Then the Gabor feature vectors of training samples and testing samples are projected to the low rank subspace through the transformation matrix. Finally, we use sparse representation classification algorithm to perform classification and recognition in the low rank subspace. Experiments on the Extend Yale B and AR databases show that the proposed method has a high recognition rate and strong robustness.