中文摘要：

人脸识别一般都要先对人脸特征做维数约简,再做识别。有些传统的维数约简算法对训练样本的数量有一定的要求,比如对分类比较有效的LDA算法。而现实应用中,数据库往往只能为每个人脸对象提供数量非常有限的图片,甚至是单样本。提出一种基于均匀LBP（Local Binary Pattern）算子和稀疏编码的人脸识别方法,使用少量关键特征代替维数约简过程,解决训练样本稀少的问题。在Stirling人脸库上进行测试,获得较高的识别率和鲁棒性,证实了算法的有效性。

英文摘要：

Face recognition generally requires facial feature dimensionality reduction before recognition. However, for some traditional dimensionality reduction algorithms, they have certain requirements on the number of training samples, such as LDA （linear discriminant analysis）, though it is quite effective in categorisation, whereas in real-world applications, often the databases can only provide very limited number of pictures for each human face, or even a single training sample. This paper presents a face recognition method, it is based on the uniform LBP operator and sparse coding, and uses few key features to replace the dimensionality reduction process, thus overcomes the problem of limited number of training samples. The method is tested on Stirling face database and achieves higher recognition rate and robustness, this confirms the effectiveness of the algorithm.