一种用于人脸识别的监督局部线性嵌入算法及其改进

摘要

为了提高人脸识别算法的识别率,提出一种基于监督局部线性嵌入SLLE(Supervised Locally Linear Embedding)的人脸图像识别方法.对局部线性嵌入LLE(Locally Linear Embedding)算法进行改进:①计算低维嵌入时,给稀疏矩阵M先加上一个单位阵,然后再计算它的特征值和特征向量,较好地解决了矩阵奇异问题；②针对LLE算法非监督的缺陷,在构造邻域的时候,增加数据的类别信息,根据其所属类别来判断样本的近邻.在Yale和ORL人脸库上的实验结果表明,该算法能够有效地提高人脸识别的性能.%In order to improve the recognition rate of face recognition algorithm, a novel face recognition method based on supervised locally linear embedding is proposed. The improvement on locally linear embedding ( LLE) algorithm is made: ① While calculating the low-dimensional embedding, the sparse matrix M is to be added a unit matrix first, then its eigenvalues and eigenvectors is calculated, this well solves the problem of matrix singular; ② For the defects of non-supervision the LLE algorithm has, the data category information is increased when constructing the neighbourhood, and the adjacent neighbours of the sample is discriminated according to their category. The experimental results on Yale and ORL face database demonstrate that the proposed method can effectively improve the performance of face recognition.