Development Of Multi-channel ECG Algorithm And Application System

The main reason cause of human death is cardiovascular diseases,ECG examination is one of the most common method of cardiovascular diseases prevention and diagnosis.In order to reduce the doctor's workload and improve the accuracy of the doctor's diagnosis,ECG signal automatic analysis technology is applied to clinical electrocardiogram diagnosis.The accuracy of ECG automatic analysis algorithm for the diagnosis of certain diseases is high.But there are still a lot of problems,such as noise and artifacts,accurate classification and intelligent multi-guided analysis.Based on the status,this dissertation carries on the study in four main aspects,i.e.waveform detection and fiducial points location,digital features and morphological features extraction,Classification and diagnosis of ECG signals.Waveform detection and fiducial points location: It is the basis of ECG automatic analysis algorithm.A new method for QRS detection was proposed by using mathematical morphology and Pan & Tompkin algorithm.The main advantage of this method is that it suppresses the influence of high T wave,baseline drift and high frequency noise on QRS complex detection.A new method for QRS complex onset and offset location was proposed based on the fusion of multi-channel ECG signals and a moving smoothing process,was called NLPD algorithm.This method avoids the localization of the peaks corresponding to QRS complex onset and offset,and improves the accuracy of the algorithm to the abnormal QRS complex.This paper first integrated traditional low-pass difference method,trapezoidal area method and clinician commonly used tangent method for P and T wave detection.First of all,the advantage of this method is to reduce the measurement error caused by P and T wave fusion,QRS wave and P wave fusion,u wave and other abnormal circumstances.Secondly,this method reduces the P-wave missed detection caused by the long PR interval.Digital feature and morphological feature extraction: It is the key to ECG diagnostic analysis.In this paper,we compare the commonly used parameter analysis methods,including: time domain,frequency domain,time-frequency domain,high-order statistical analysis.In this paper,the basic morphology of the QRS complex is classified by the method of the Sparse Auto-encoder(SAE).Classification and diagnosis of ECG signals: A novel approach for PVC/N heartbeat classification was proposed based on ECG features and Support Vector Machine(SVM).The parameters selected by this algorithm are based on the parameters proposed in this paper,such as Max Amp All Samp,and other time domain QRS complex parameters.Other types of arrhythmia Classification can be carried out on this basis.In addition,there are few studies on arrhythmia analysis based on the depth of learning,we use the Sparse Auto-encoder(SAE)to divide the ECG heartbeat into PVC / N.The Validity of the methods mentioned above has all been proved using MIT-BIH database.In order to validate and test the above algorithm,this paper builds up the ECG signal acquisition platform and collected 600 cases of different age,gender,disease type of clinical 12 lead ECG signal.After a doctor's calibration,a Chinese-based ECG database was created,called SZU database.The database is more suitable for ECG automatic analysis algorithms for Chinese people testing and development.