Research Of ECG Signals Of Arrhythmia Based On Tensor Analysis

Arrhythmia is the abnormality of the frequency or rhythm of heart beat due to the origin or conduction of cardiac activity.The incidence of arrhythmia is high,which seriously endangers the health of the patient.The malignant arrhythmia will lead to sudden cardiac death and other undesirable consequences.The rapid development of ECG processing and ECG signal extraction as the detection means provide the basis for the study of arrhythmia.Arrhythmia ECG contains a wealth of time-varying information.In recent years,tensor analysis has attracted increasing attention in pattern recognition and feature extraction.Tensor can be seen as a vector to the high-dimensional space on the expansion,and will not destroy the intrinsic relationship between the elements.Based on the above characteristics,this paper uses the time-frequency analysis method to elevate the single-conductor electrical signal in the observation window as a high-dimensional tensor.Using the tensor analysis method to extract the characteristics of ECG signal of arrhythmia,can effectively use the ECG signal self-similarity,synergy and structural information.This paper mainly proposes two ECG feature extraction algorithms based on tensor analysis:The one,ECG classification algorithm based on wavelet-tensorization and Tucker decomposition: In this paper,the ECG signal is divided into frames,the third order tensor is constructed by wavelet decomposition,and the projection vectors in three directions are obtained by Tucker decomposition.Using support vector machine(SVM)to complete the classification,and selecting the better classification of the projection direction and framing strategy;The other one,ECG classification algorithm based on FRFT tensorization and non-negative tensor decomposition: In this paper,the ECG signal is divided into frames,and the three-order tensor is constructed by the fractional order Fourier transform.Based on the core tensor and the projection matrix getting from non-negative tensor decomposition to calculate the tensor Frobenius norm of training set and testing set.This paper uses the Physio Net open database to conduct the simulation experiment,proving that tensor analysis can effectively distinguish atrial fibrillation,ventricular fibrillation,chamber speed and normal ECG signal.This paper explores the application of tensor analysis in electrocardiogram and provides a new type of arrhythmia detection method and index,which can be used in clinical diagnosis,guardianship,the equipment of wearable health management,etc.