中文摘要：

基于滑窗QR分解不但能快速、准确地更新正交投影,同时还可提供其协方差的rank-k更新表达,提出了一个用于信号子空间更新的快速PCA算法.通过对进行主元分析（PCA）计算的非线性迭代部分最小二乘算法（Non-linear Iterative Partial Least Squares,NIPALS）计算过程的改进,将特征向量的更新转化为小维度辅助向量的更新,在满足特征值和特征向量更新精度的同时,有效地提高了计算速度.将滑窗QR和快速PCA算法用于子空间辨识算法的自适应更新,数值仿真验证了此自适应子空间辨识算法的有效性.

英文摘要：

Moving window QR can not only update the orthogonal projection rapidly,but also describe it as a rank-k modification problem.Based on it,a rapid principal component analysis（PCA）algorithm with high accuracy was proposed to update the signal subspace accurately,which accelerated the non-linear iterative partial least squares（NIPALS）by transforming the computation of eigenvector into the updating of three small-size vectors.The proposed adaptive subspace identification algorithm combines moving window QR and rapid PCA technolgies together.Numerical simulation study demonstrates the efficiency of the technology.