ECG Signal Processing Algorithm And Its Implementation For Wearable Applications

Cardiovascular disease is one of the diseases with the highest morbidity and mortality in the world,and it has seriously threatened people’s health.Electrocardiogram（ECG）is a direct reflection of human heart electrical activity,and it’s an important basis for the diagnosis of cardiovascular disease.The detection steps of traditional 12-lead ECG monitor are complicated and it requires professional medical staff to operate.It cannot realize real-time feedback of ECG activities,which makes some sudden heart diseases not timely diagnosed and treated,which greatly threatens people’s lives.With the development of wearable ECG monitoring technology,it is particularly important to efficiently process ECG signals in wearable application scenarios.In this thesis,a fusion feature extraction algorithm combining wavelet packet decomposition-statistical analysis and slope threshold is proposed.First,the wavelet packet function is used to decompose the ECG signal into four layers to obtain 16wavelet packet decomposition coefficients.The singular value,maximum value,and standard deviation are calculated to obtain frequency-domain feature F_{（540×48）}.At the same time,the slope threshold method is applied to detect R-peak on the de-noised ECG signals and calculate the RR interval.The first two RR intervals are extracted as time-domain features T_（540×2）.Finally,the obtained feature space in the frequency-domain and time-domain are employed to form a fusion feature space M_{（540×50）}.According to different types of ECG signals,a combination of waveform morphology analysis（WMA）and support vector machine（SVM）is chosen to classify ECG signals.9 types of ECG signals need to be classified,which include:normal ECG（N）,sinus bradycardia（SB）,ventricular tachycardia（VT）,premature ventricular contraction（V）,premature atrial contraction（A）,atrial fibrillation（AF）,atrial tachycardia（AT）,sinus arrest（SA）,sinus tachycardia（ST）,respectively from the MIT-BIH database,LTAFDB database and Fluke Physiological parameter simulator.Because SB,A,AT,SA,and ST have obvious characteristics in rhythm,these five types signals are classified by WMA.The SVM are used to classify the other four types signals.Through simulation analysis of MATLAB,the accuracy of the combined classification method is 96.67%.Finally,the ECG signal front-end acquisition circuit is designed,Tencent Cloud is performed to build a cloud server,and a mobile（Huawei Honor 8:FRD-AL10）APP is applied to realize human-computer interaction.The wearable-oriented ECG signals processing algorithm proposed in this thesis can effectively extract the characteristics of ECG signals,and can classify and diagnose human signals collected by wearable devices through a combination of WMA and SVM.In addition,the proposed algorithm can be used in actual wearable devices,which has important practical significance for the prevention and pre-diagnosis of heart disease monitoring and treatment.