Feature Extraction Of Multivariate Hysteretic Time Series

For now,heart disease is the first cause of death in the world and the risk of death and morbidity of heart disease in China tends to be younger and more likely to rise,so it's important to study the information about the heart.In this study,we found that there is a hysteresis loop in the process of electrocardiogram transmission in the human body with stimulation signals and unstimulated signals.At present,the research on the pattern recognition of multiple hysteresis time series is still blank.In order to study the hysteresis transmission characteristics of multiple hysteresis time series in human body meridian,this paper analyzes the human body meridian communication,and the main contents are as follows:(1)In order to study the meridian acupoints multivariate hysteresis ECG signal,this paper designed a noninvasive meridian electrocardiosignal acquisition experiments.Using two cascade RM6280 C multi-channel physiological instrument,the electrodes were attached to the nine points on the meridian of the hand of the left arm of the human body,and the human body meridian signals were collected.It is a good filter for the signal to be used to get the signal,to use the high-pass,low pass,to remove the resistance of the 50 Hz operator and the default threshold,which is to be used to filter out the noise,and to retain the useful signal,it has very good filtering properties.And then we divide the main wave of the QRS,and extract the biggest lag ring in the electrical signal.(2)Based on the multivariate hysteresis characteristics of the human body's meridian signals,this paper presents the modeling of the multiple hysteresis time series based on the Duhem model.Because the Duhem formula is explicit and applied to dynamic nonlinear systems,it's a good idea to use the Duhem model to model the human body's meridian signals.When choosing model order,using the first approximation theorem and generalized Weiestrass recursive least squares algorithm as the Duhem model parameter identification method.Experimental results show that the original output is consistent with the model output and the modeling results are good.(3)In order to study whether there are characteristics or generality of the human body meridian signals,this paper proposes a multivariate delay time series feature extraction method based on the Duhem model,and uses the model structure parameters as the characteristic data.In order to verify the validity of the method based on the Duhem model,two data sets are used to verify the validity of the method.One is the human body meridian signal with multiple delay time series of meridian points and the other is the data of the vehicle damper.Both data sets were extracted with Duhem model features,and the identified model parameters were used as the eigenvalues of the data.Then we use the k-means,FCM,and the hierarchical cluster to analyze the total eigenvalues matrix,and use the method of PCA to do a control to verify the validity of the different features extraction methods.According to the experiment,based on the Duhem model feature extraction method,the data clustering RI of the vehicle damper is 20 percentage points higher than that of the PCA feature extraction,and it is 3 percentage points lower than that of the human body meridian data clustering.It was found that that hysteresis of the electrocardiogram signal is small,and the hysteresis loop of the data of the automotive damper is particularly large,which show that the method based on the model feature of the Duhem model is more applicable to the data of the hysteresis characteristic.(4)In order to explore the more effective method of the delayed characteristic of the electroshock signal,this paper proposes to extract the characteristics of the sequence of the meridians in the temporal sequence of the transfer entropy.It is found by calculate that transfer entropy value between points of acupuncture point that there is a certain correlation between the transfer entropy value and the size of the hysteresis loop of the cardiac signal: For different people,the cardioelectric signal is a hysteresis large ring and the entropy value is low;Ecg signal is small hysteresis loop,transfer entropy is high.The transfer entropy is small;The electrocardiogram is a hysteresis loop with a high entropy transfer.In order to prove the validity of the entropy feature extraction method based on relay,PCA feature extraction method is used as a control group,in accordance with the hysteresis characteristics of different people and in accordance with the size of the hysteresis loop respectively through cluster analysis,experiments show that the feature extraction method based on the transfer entropy clustering results are better than PCA.The results of the full-text experiment can be used to obtain the effect of the feature extraction of the multiple hysteresis time series of meridian points: Based on transfer entropy is better than PCA and based on the Duhem model.