生理信号时间序列周期性和平稳性对近似熵和样本熵算法的影响分析

摘要

目的提高近似熵和样本熵算法在评价生理信号时间序列非线性复杂度应用中的精度.方法 首先生成生理信号时间序列数据库,通过对周期序列和叠加有周期成分的非线性序列的分析,研究序列周期性对熵测度算法的影响,并通过对心率变异性(heart rate variability,HRV)序列在去除非平稳趋势前后的对比分析,研究序列平稳性对熵测度算法的影响.结果 在序列长度范围内,不同重复频率的周期序列熵测度不同,不同比重的周期成分叠加到非线性序列中引起序列熵测度的变化也不同.生理信号时间序列中大都存在非平稳成分,而非平稳成分会降低序列的复杂度,因此进行熵测度计算前首先要去除非平稳成分.结论 周期性和非平稳成分显著影响生理信号时间序列的熵测度算法.%Objective To improve the accuracy of approximate entropy (ApEn) and sample entropy (SamEn) in evaluating the nonlinear complexity of physiological time-series. Methods We firstly constructed a database of physiological time-series. Secondly, we studied the influence of periodicity on entropy measure by analyzing the periodic sequences and nonlinear sequences added with different proportion of periodic components. Thirdly, we investigated the influence of stability on entropy measure by comparing between the original heart rate variability (HRV) sequences and the HRV sequences after detrending. Results Within the limited length of time-series, the periodic sequences with different repetition rate had different calculated results of ApEn and SamEn, and different proportion of periodic components added to the nonlinear sequences could change the entropy measure. As physiological time-series were usually with some non-stationary components which reduced the complexity of the sequences, the non-stationary components should to be eliminated before the calculation of ApEn and SamEn. Conclusions The existent periodic and non-stationary components significantly affected the entropy measures.