ECG Arrhythmia Feature Extraction And Analysis By Multi-scale Time-frequency Method

Arrhythmia is an important group of diseases among cardiovascular disease, which is a kind of heart beat frequency or rhythm abnormal caused by the origin or conduction disturbances of heart activities. And may lead to sudden death, also it can cause heart failure because of persistent accumulation. ECG is a bioelectric signal generated by the excitement in cardiac activity, which is an important objective index in the analysis of the generation, conduction and recovery process of cardiac stimulation.Heart rate variability(HRV) refers to the minute fluctuation between successive RR intervals, which contains a lot of information about the regulation of neural and humor on cardiovascular system. And the complexity of HRV signals can reflect the heart physiological function and health status. Automatically identify and distinguish different types of arrhythmias by analyzing HRV signals show great importance to clinical diagnosis and remote monitoring technologies. Commonly used analysis methods of HRV signals include time domain, frequency domain and nonlinear methods. Among thesee, the nonlinear analysis method is widely considered to be more helpful to reveal the essential characteristics of the power system of human heart.In this paper, HRV signals were extracted from ECG records after the signal preprocessing, and feature extraction and analysis were conducted on different aspects, such as multiscale entropy, scatterplots and symbol dynamics.This paper firstly presented a multi-scale Renyi entropy algorithm and made atrial fibrillation(AF) feature extraction and analysis as an example, the result of paroxysmal atrial fibrillation classification can reach more than 92%. Secondly, this paper presented an improved scatter plot area entropy algorithm, which can reflect the nonlinear characteristics and complexity changes of ECG signals when heart disease occurs. The the correct rate of AF classification can reach 90% or above. In addition, this paper also presented a scatterplot region encoding method. Under the new coding method, this paper presented a symbol sequence entropy algorithm based on the scatterplots, which could reflect the different change patterns of RR intervals from scatterplot level and the experiment with different arrhythmia signals and normal sinus signals verified the effectiveness of this algorithm, which expanded the application of symbolic dynamics in ECG analysis.