Research On Nonlinear Time Series Analysis Method Based On Statistical Properties Of Complex Network

Some biomedical signals have nonlinear properties,so they belong to nonlinear time series.As for modern signal processing,biomedical signal is an important field,and it mainly contains three categories: electroencephalograph(EEG),electrocardiogram(ECG)and electromyography(EMG).The research of medical signal is helpful to the development of medical treatment and is of great significance to improve the level of human health.The number of cardiac patients and epilepsy patients increases gradually in recent years,which affects the patients' normal life seriously.Medical studies have shown that EEG and ECG signals have nonlinear properties,which reflect the electrical activity of brain and heart respectively and become the scientific basis for the diagnosis and treatment of patients.The emergence of construction algorithm makes it possible for researchers to analyze the nonlinear time series by using the statistical properties of complex network.This paper analyzes the epileptic EEG,ventricular fibrillation(VF)and ventricular tachycardia(VT)signals and proposes new classification algorithms respectively to improve the classification performance.As for epileptic EEG,the classification algorithms in this paper can distinguish the interictal and ictal EEG accurately and take scientific medication and treatment according to the different conditions.For cardiac patients,the main cause of sudden cardiac death is the deterioration of VF or VT.The treatment measures for the two symptoms are different;if we misjudge them,it will cause irreparable harm to the patients.As for VF and VT,the classification algorithm in this paper can distinguish the VF and VT accurately and take treatment for patients timely,which is of great significance to reduce the cases of sudden cardiac death.Firstly,we transform the experimental data from time domain into complex network domain by utilizing the horizontal visibility graph algorithm.In this algorithm,each sampling point of time series is considered as a node in complex network and whether there is a connection between nodes depends on the local convexity property.This algorithm does not involve any parameter in the process of constructing complex network,which reduces the subjectivity greatly,and gets the complex network corresponding to the time series precisely.Secondly,we analyze the structure of complex network and its statistical properties.Asfor epileptic EEG,we extract the degree centrality and its mathematical transform as the classification features to classify the epileptic EEG,and the classification results are better than the traditional nonlinear analysis methods,such as sample entropy and approximate entropy and so on.Considering the linear features also can reflect some information of time series,we extract the degree centrality combined with fluctuation index and variation coefficient as three-dimensional vector to classify the data,and the classification performance is better than single feature.Through further analysis of the complex network,we construct new features and the sum of features to classify the epileptic EEG,and the classification performance can be raised to a higher level,which has important significance in diagnosis and treatment of epilepsy.At the same time,we utilize the complex network theory to study the VF and VT for the first time,which provides a new method to study ECG signal.Through the analysis of complex networks corresponding to the VF and VT,we propose a new classification algorithm based on degree centrality,and the classification result is much better than the traditional nonlinear analysis methods and the complexity algorithm,which has better classification performance,and the classification accuracy is up to 99.5%.The experimental results show that the complex network theory is suitable for the study of EEG and ECG signal,and the statistical properties of complex network can reflect the nonlinear dynamic information of the original time series.The new classification algorithms we proposed in this paper can identify different medical signals accurately,which is helpful to the diagnosis and treatment for epilepsy patients and cardiac patients.The research ideas and classification algorithms in this paper are of great significance to the study of medical signals and contribute to improve the medical level.