基于平滑先验法的两种熵研究

Two Entropy Studies Based on Smooth Prior Approach

摘要: 从心电系统中采集的心率变异性(Heart Rate Variability, HRV)信号，不可避免的混入噪声和各种趋势，消除这些趋势的干扰，有助于非线性系统复杂动力学分析的准确性。本文采用平滑先验法消除健康人白天和夜间的HRV信号中所叠加的不同趋势项，使用近似熵和基本尺度熵对其进行复杂性计算，对比分析去趋势处理前、后近似熵和基本尺度熵的稳定性变化。研究发现：去趋势处理后，平滑先验法能够有效去除HRV信号的各种趋势项，且较大幅度地提高了近似熵的稳定性，而基本尺度熵在去趋势处理前、后均呈现出较强的稳定性。研究结果表明去趋势处理后，两种熵测度都可以区分开健康人白天和夜间的HRV信号，充分体现出平滑先验法可以提高熵测度的稳定性和识别性，为临床的生理病理诊断提供了重要的理论依据。

Abstract: The heart rate variability (HRV) signals collected from the ECG system are affected with noise and various trends inevitably. Eliminating the interference of these trends contributes to the accuracy of complex dynamic analysis of these nonlinear systems. In this paper, we use the smoothness prior method to eliminate the different trends superimposed in the HRV signals of healthy people during the day and night, and use the approximate entropy and the basic scale entropy to calculate the complexity. Before and after removing the trending interference, we contrast and analyze the stability changes of approximate entropy and base-scale entropy. The results showed that after detrending processing, the smooth prior method can effectively remove various trend interference from the recordings, and can greatly improve the stability of approximate entropy. The experimental results also showed that the base-scale entropy indicates strong stability regardless of whether or not to use detrending processing. After detrending, the two entropy measures can distinguish the HRV signals of healthy people during the day and night. It fully reflects that the smoothness prior method can improve the stability and recognition of the entropy measure, and provide an important theoretical basis for clinical physiology and pathological diagnosis.