改进的周期信号盲抽取算法①.PDF

计 算 机 系 统 应 用 2014 年 第 23 卷 第 6 期 改进的周期信号盲抽取算法① 杜 瑞, 王剑书 (西北工业大学 电子信息学院, 西安 710072) 摘 要: 周期信号的盲抽取在医学等领域具有重要的研究价值. 对于此类信号的抽取需要利用先验知识对其周 期进行较为准确的估计, 多数算法对于目标信号周期估计误差十分敏感. 文章针对此问题提出一种改进的周期 特征信号盲抽取算法, 首先通过先验知识估算最佳时延的存在区间并利用改进的时域相关函数对提取矩阵进行 初始估计, 再使用负熵作为目标函数并引入自相关函数确保抽取顺序, 使用牛顿迭代法进行优化, 达到最优结 果. 仿真和实际 ECG 数据测试表明, 对比几种现行算法本文提出的算法抽取效果较好, 并增强了周期估计误差 的鲁棒性. 关键词: 盲抽取; 时域相关函数; 负熵; 牛顿迭代法 Improved Algorithm for Extracting Periodic Signals DU Rui, WANG Jian-Shu (School of Electronic and Information, Northwestern Polytechnical University, Xi’an 710072, China) Abstract: The research of extracting periodic signals is of great value in the field such as medicine. A precise estimation for period of target signals with priori knowledge is necessary in such kind of algorithms. But most of algorithms are sensitive to the error from the estimation. The algorithm proposed in the article including steps as follows: first roughly evaluate the extracting matrix with the improved time domain correlation function by evaluating the existing interval of optimizing time delay, use negentropy as the target function with the correlation function given into consideration to ensure the order and optimize it with Newton iterative method. The simulation and the experiment with real ECG data prove that the proposed algorithm can get a better result in the extraction than other algorithms, and show its robustness to the error of time delay. Keywords: BSE; time domain correlation function; negentropy; Newton iterative method 盲抽取(Blind Source Extraction, BSE)是一种利用 Barros A K 等[1]在 2001 年提出了一种对感兴趣准 先验知识从混合信号中抽取一个或者一部分感兴趣信 周期信号的快速抽取算法, 但该算法高度依赖于周期 号的方法. 相比盲分离(Blind Source Separation, BSS) 估计的精度. Zibulevsky M 等[2]在 2002 年提出一种利 方法, BSE 在提取感兴趣信号时具有明显优