一种改进的混沌序列去噪方法-太赫兹科学与电子信息学报.PDF

第 9 卷 第 5 期 信 息 与 电 子 工 程 Vo1．9，No．5 2011 年 10 月 INFORMATION AND ELECTRONIC ENGINEERING Oct．，2011 文章编号： 1672-2892(2011)05-0586-05 一种改进的混沌序列去噪方法 韩贵丞，李 锋 (复旦大学 信息科学与工程学院，上海 200433) 摘 要 ：针对混沌时间序列的噪声平滑，分析了其信噪比与关联维数的关系，并结合局部投 影方法的邻域选取问题，提出了一种改进的去噪方法。该方法对于含有噪声的混沌信号选定一个 邻域半径值，计算去噪后序列的关联维数，确定该选定的邻域半径值是否准确并进行调整，最终 找到合适值并进行多次局部投影去噪，从而得到去噪后的纯净序列。用改进后的方法对含噪声的 Lorenz 序列和对股市序列进行对比预测仿真，仿真结果表明该方法能够有效地选取邻域半径，进而 改善局部投影方法的去噪效果，取得更好的预测效果。 关键词 ：混沌时间序列；局部投影算法；关联维数 中图分类号 ：TN919.5 ；O415.5 文献标识码 ：A An improved algorithm for de-noising of chaotic data HAN Gui-cheng，LI Feng (Department of Electronic Engineering，Fudan University，Shanghai 200433，China) Abstract： For the denoising of chaotic time series, the relationship between correlation dimension and SNR was analyzed. Combined with the neighborhood selection of local projection method, an improved method of noise reduction was proposed. This method selected a neighborhood radius for chaotic signal with noise and calculated the correlation dimension to determine whether the value of neighborhood radius should be adjusted after de-noising. Eventually, the appropriate value was found to iteratively calculate with the local projection method to get the pure de-noised sequence. After simulating the improved method in the noisy Lorenz series and stock index series, the results show that this method can effectively select the radius of the neighborhood, thereby improve the local projection method in noise reduction