Statistical Complexity Measureanalysis Of ECG Signal Based On LMCD And JSD

As an important index to describe the stochastic nature of sery, statistical complexity measure played more important role in analyzing time series and symbolic sequence. In this thesis, two different statistical complexity measures were used to analyze and to research two cardiac arrhythmia signals compared with normal sinus rhythm.Firstly, basic knowledge of ECG signal and statistical complexity measure were introduced and then statistical complexity measures of two typical cardiac arrhythmia signals and normal sinus rhythm were computed and analyzed on arrhythmic ECG.Secondly, two different divergence measures were introduced: Lopez-Mancini-Calbet Divergence (LMCD) and Jensen-Shannon Divergence (JSD). They both could quantize the divergence of probability distributions. LMCD was defined in the Euclidean space and statistical complexity measure based on LMCD was hard to reflect on the stochastic nature and structure information because it was just a simple form of entropy density. JSD did not require the condition of absolute continuity of the probability distributions. More important, it had close relationship with the variational distance. Statistical complexity measure based on JSD not only could reflect the structure and detail information of time series, but also could quantize its stochastic property well.Finally, statistical complexity measures based on LMCD and JSD were used to compute complexities of NSR, CHF and SCD. Bandt-Pompe algorithm was used to compute pattern probability of symbolic sequences. Three ECG signals were choosed 1000 points(7 cycles), 5000 points(36 cycles) and 10000 points(72 cycles) respectively to compute statistical complexity measures.The results were projected to Complexity-Entropy plane.All the results shown that three ECG signals had different statistical complexity measures and different projection areas. NSR has the highest statistical complexity, followed by CHF, and SCD has the lowest statistical complexity. The difference of ECG signals'projections and their areas would become larger while the length of data (number of cycle) growing. The variance test indicated that statistical complexity measures based on LMCD and JSD could disclose significant differences among these three signals'complexities.But statistical complexity measure based on JSD shown better differentiation.