Research On ECG Information Collection And Arrhythmia Detection Methods

The discovery and identification of heart disease is mainly achieved by analyzing and interpreting the waveform characteristics of the electrocardiogram.The electrocardiogram is generated by measuring,collecting,processing and displaying the human body’s ECG signal using an ECG detection system.The current ECG detection systems commonly used in hospitals mainly include large electrocardiographs and small dynamic ECG monitoring devices(such as Holter)that can be worn by them.These devices generate an electrocardiogram by measuring the patient’s ECG signal and are diagnosed by the physician based on medical knowledge and experience,with important clinical value.However,there are also deficiencies such as high price,complicated operation,and must be operated and analyzed by professional medical personnel to obtain diagnosis results,which is not convenient for daily household testing.With the rapid development of machine learning and signal processing technology,the realization of an ECG signal detection system that is easy to wear,reasonably priced,and has an automatic diagnostic function becomes possible.Firstly,according to the principle of ECG signal generation and measuremen t,with the BMD101 ECG acquisition module as the core,a chest-mounted wearable ECG acquisition and transmission terminal with low price,small size and simple operation was designed.The terminal can transmit the collected ECG signal to the PC through the Bluetooth transmission protocol,and the PC can draw the ECG waveform waveform in real time,and save and generate the txt format ECG data file,which can be read and drawn by MATLAB for subsequent Process analysis.Then,the ECG signal is filtered and denoised and waveform detected to obtain the main waveform characteristics of the ECG signal.Taking the ECG signal in the MIT-BIH arrhythmia database as an experimental sample,the band-pass filter and wavelet threshold denoising method are flexibly applied to the amplitude-frequency characteristics of the ECG signal and its main noise source,and the parameters are reasonably set.The denoising effect of the ECG signal is improved by using an improved differential adaptive threshold detection algorithm to improve the detection accuracy of the R wave and QRS complex.Finally,according to the waveform detection results of ECG signals,ECG data calculation and arrhythmia classification are realized.On the one hand,heart rate,heart rate variability,body fatigue and mental fatigue were calculated according to the R wave interval,and the determination of five common arrhythmia conditions was completed.On the other hand,read the various arrhythmia labels indicated in the annotation files of the MIT-BIH arrhythmia database,and extract the instantaneous RR interval,relative RR interval and 45-dimensional QRS complex eigenvalues of various ECG beats.After normalization,a 47-dimensional feature vector data set space is generated.Gaussian kernel SVM and CNN classification models were used for classification experiments,and all of them achieved good test results.The overall classification accuracy rate was over 90%.However,by comparison,the Gaussian nuclear SVM model has slightly better test data than the CNN model,and is more suitable for arrhythmia classification.