Assessment Of ECG Signal Quality Based On Multi-Feature Fusion

Cardiovascular disease(CVD)is one of the main causes that threatening human’s life and health,with high morbidity and mortality.It has become a worldwide problemthat how to strengthen the prevention and treatment of cardiovascular disease.Electrocardiogram(ECG)characterizes the changes in myocardial potential during the process of cardiac rhythmic contraction and relaxation,which contains a large number of information related to disease and physiology.Analysis of ECG signal has become the key to the prevention and treatment of cardiovascular disease.With the development of mobile devices and mobile Internet technology,portable ECG signal acquisition devices are used to collect ECG signal.As the human body is a complex and changeable environment,the signals that we collected are susceptible to physical and external interference,which resulting in loss of clinical significance of ECG signal,increasing the workload of doctors and misdiagnosis rate.Therefore,before ECG can be diagnosed,it should have an assessment to evaluate its quality and classify it as two categories of quality and quality of failure.In this paper,the physiological characteristics of ECG signals and the interference are analyzed in detail,and the self-designed ECG signal acquisition device is introduced.The differences between the quality and the quality failure of ECG signals are discussed.Thus,the threshold classification method and the machine learning method are used to classify the ECG signals based on multi-feature fusion.The main contents are as follows:1.ECG signal acquisition deviceThe design uses AD8232 as the acquisition chip,STM32 as the main control chip,the collected ECG data will be transfered to the computer,and for further analysis,thus more intuitive and specific understanding of characteristics and acquisition process of ECG signal.2.Threshold discrimination method for ECG signal qualityThrough the ECG signal’s lead-off,maximum voltage,baseline drift,kurtosis and the correlation of different lead signals’,several experiments are made according to the physiological knowledge of ECG signal.The ECG signals which thourgh selections areunqualified signal.Then continue typing other characteristics of the remaining signal threshold discrimination,til the ECG signals are classify into two types.3.Machine learning method for ECG signal qualityThis paper not only analyzes the characteristics of time domain waveforms and frequency domain energy of ECG signals,but also analyzes the nonlinear space of ECG signals just like multi-scale entropy of signals with a view to more comprehensive information on ECG signals.The ECG signal characteristics combined with the machine learning method will be extracted to classify and qualitify the signal.