关键词: 子空间投影算法, 特征分解, 增量主分量分析, 近似协方差矩阵

Abstract:

Firstly, with eigenvectors orthogonal to each other, the computation complexity of the subspace projection(SP) algorithm is reduced to 1/P of the original algorithm(where P is the number of desired eigencomponents). Then, the covariance matrix is replaced by the approximated covariance matrix which is composed of large eigenvalues and corresponding eigenvectors, the computation complexity can be reduced to 1/N of the original algorithm(where N is the input vector dimension)further. Finally, experimental results based on the ORL face database demonstrate the efficiency of the presented algorithm.

Key words: subspace projection, eigenvalue decomposition, incremental principal component analysis, approximated covariance matrix