基于核贝叶斯压缩感知的人脸识别

为加快人脸识别速度和提高人脸识别率,将贝叶斯压缩感知算法进行核扩展并运用到人脸识别,改进局部特征统计方法,结合空间金字塔模型,用于人脸图像的特征提取。首先用局部特征统计提取图像特征,在此基础上再进行第二层局部统计,然后根据空间金字塔模型分层提取不同空间尺度的特征,最后运用核贝叶斯压缩感知算法分类。在AR和FERET人脸数据库上的试验结果表明,本研究算法相对于传统方法具有更好的性能。

Face recognition based on kernel Bayesian compressive sensing

In order to improve the speed and rate of face recognition, Bayesian compressive sensing algorithm was applied and its kernel extension to face recognition was proposed. Combined with the spatial pyramid model, statistical local feature was improved to extract the features of face images. Firstly, the statistical local feature was used as a feature extractor to obtain facial features and a second layer of local statistics was processed based on the former layer. Then the spatial pyramid was used to obtain features in different spatial scales in order to accomplish the final step of face recognition, the features were classified through kernel Bayesian compressive sensing. The experimental results on the basis of the AR and FERET databases demonstrated that this algorithm had better performance than other traditional ones.