基于RLS的特征提取算法研究

The Research of Feature Extraction Algorithm Based on RLS

针对主成分分析在提取特征脸时需要计算各个特征值和对应的特征向量,而导致的计算量大的问题,提出了一种基于递归最小二乘(RLS)的特征提取算法。该算法每迭代一次,只需计算一个特征值和其对应的特征向量,使得计算量大大减少。实验表明:在一定迭代次数内,选取的主成分越多,RLS算法得到的图像的重构率就越大。

Aiming at the problem of PCA when it is used to calculate various Eigen values and the corresponding eigenvectors which may led to large computational problems,we study the recursive least squares(RLS) method to extract the features.For each iteration,we just to calculate one Eigen value and it corresponding eigenvector, so the calculation will be greatly reduced.The experimental result shows that in a certain number of iterations, the more main component we select,the larger reconstructed rate of face we will get when use the RLS algorithm.