一种伪周期子序列精确匹配算法的研究

摘要

伪周期序列中各个子序列的形态变化,蕴含了丰富的动态特性.通过匹配算法,比较子序列的形态差异,是建立系统动态模型的基础.SEA(shape exchange algorithm)方法是一种新颖的伪周期序列匹配算法,能够克服不同序列的时间和空间尺度差异,对噪声和偏移量具有很好的鲁棒性.对于同一时间序列中不同子序列的匹配问题,则缺乏对幅度差异的描述能力.鉴于SEA方法对细微形状畸变的敏感性,采取了幅度特征互换的方法,人为引入形状畸变,提高算法对幅度差异的刻画能力.为了不损害SEA方法的鲁棒性,选用中值作为待交换的幅度特征,并从中扣除偏移量.以心电信号为例,通过实例检验证实,修改后的SEA方法适合处理子序列匹配问题,并仍具有很好的鲁棒性.%In pseudoperiodic series, the morphology of sub-series contains rich dynamic characteristics. Comparing with match algorithms, the morphology difference of sub-series is the premise for constructing system dynamical model. Shape exchange algorithm (SEA) is a novel pseudo periodic series matching algorithm. It can overcome the space and time scale difference of the sub-series, so it is very robust to noise and offset. SEA can not describe the amplitude difference when matching the sub-series of the same time series. A strategy of amplitude feature exchange is used to import artificial shape deviation for SEA, which is sensitive to little shape deviation. The strategy can improve the describing ability for amplitude difference remarkably. The media value is selected as an amplitude feature to protect the robustness of SEA and the offset component is eliminated beforehand. Taking ECG as an example, it is demonstrated that the modified SEA is suitable to processing sub-seties matching and still maintains good robustness.