小波互信息及其在心电分析中的应用

Wavelet Mutual Information and ECG Analysis

为探究信号小波成分中的有用信息,提出了从小波变换(WT)系数和多层小波分解(WD)成分的延时互信息(MI)得出信号分析指标的新方法。以年轻(21～34岁)与年老(68～81岁)二组健康人心电图(ECG)为实验数据,用Symlets 4为母小波对ECG进行WT和WD,计算出MI与心率周期对应的极大值(LM)和表示该波峰形态特征的波形宽度(WW)等多个特征指标。结果表明,在尺度a=12时的WT系数及第3层细节的LM都是年轻组明显小于年老组(P<0.001),第2层近似和第3层细节的WW都是年轻组明显大于年老组(P<0.001)。ECG的延时互信息可揭示出心率变异性随年龄而减少。

To explore the useful information in wavelet components of a signal, a new method is presented. The analysis indices of the signal are obtained from time-delayed mutual information（MI） of wavelet transform（WT） coefficients and multilevel wavelet decomposition （WD） components. The electrocardiograms （ECGs） are used as experimental data collected from young （21-34 yr） and elderly （68-81 yr） groups of healthy subjects. The data analysis is performed by WT and WD using Symlets 4 as a mother wavelet. The multiple characteristic indices of MI are calculated, such as the local maximum （LM） corresponding to the period of the heart rate, the waveform width （WW） figure feature of the wave crest, etc. Results show that both LM of WT coefficients at a = 12 and that of the detail at the third level are significantly lower in the young group than that in the elderly group （P〈0. 001）. WWs of the approximation at the second level and the detail at the third level in the young group are greater than that in the elderly group （P〈0.001）. The time-delayed MIs of ECG indicate that the heart rate variability is decreased with aging.