Research On Signal Processing And Classification Methods For Wearable ECG Monitoring Devices

Since personal health is getting more and more concerned,wearable ECG monitoring devices are booming in the market.Therefore,ECG signal processing and classification methods that are suitable for wearable ECG devices become a hot topic.In recent years,ECG signal processing and classification methods have achieved rapid development,but automatic EC G diagnosis with more efficiency and more accuracy is always pursued by researchers.Research on EC G signal processing and classification mainly includes signal p reprocessing,feature extraction and classification of arrhythmia.The various noises in EC G signal acquired by wearable ECG monitoring devices are harmful to the following analysis.Therefore,the first step of ECG signal analysis is preprocessing in order to remove noises and obtain pure signal.The thesis proposes an all-phase combination notch filter to remove the power grid frequency interference which is not at the integer frequency.Experimental results show that the combination filter is capable of removing various noises in ECG while keeps the integrity of ECG.According to the characteristics of different types of arrhythmia,the thesis analyzes the time-domain features and frequency-domain features of EC G and extract five time-domain features.Nevertheless,since EC G signal is non-linear Gaussian random signal,features in both time domain and frequency domain can not represent the non-linear information of the signal.To solve the problem,the thesis uses higher-order statistics to represent the non-linear and dynamic features of EC G signa l.The third-order accumulation of ECG are calculated and extracted as features.Finally,a support vector machine is used to classify five categories of arrhythmia,including normal cardiac beats and the result is fairly good.