摘要
提出以能量谱中的截断频率之倒数作为相空间重构过程中的窗长 ,在窗长固定情况下 ,利用奇异值分解算法确定嵌入维数和时间延迟两个参数 ,克服了不考虑窗长单独选择嵌入维数和时间延迟造成的相关维数收敛性差的缺点 ,大大提高了计算效率 .采用迭代奇异值分解算法对含噪声的信号进行降噪 ,降低了噪声对相关维数计算结果的影响 。
Correlation dimension can be estimated efficiently and accurately with the knowledge of an optimal given window length, embedding dimension and delay time. This paper investigated how to get the optimal window length, embedding dimension and delay time in order to improve the estimate of the correlation dimension. And a method using iterative singular value decomposition (SVD) was presented for reducing the noise from a nonlinear time series to yield an improvement in the correlation dimension estimation.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2002年第8期1214-1217,共4页
Journal of Shanghai Jiaotong University
