Research On The Detection And Classification Algorithm For ECG Signal

With the development of society and people’s living standards,resulting in the changes of people’s living habits,and the cardiovascular diseases has increased significantly.The heart disease is the third leading cause of death.People suffering from the heart disease usually show arrhythmia at the beginning of their illness.In order to promote the application and development of modern information processing technology in biomedical engineering,more and more research about ECG signals related to automatic detection have been carried out.However,the ECG signal exhibits low frequency,multi-noise,individual differences and adds certain difficulties to the detection of ECG signals.The current ECG signal detection algorithm is not accurate,and the paper research on ECG signal preprocessing,waveform recognition and classification algorithms was carried out.Three points of contents was summarized as shown in follow.Ⅰ.ECG signal preprocessingECG signal is a low frequency signal with many noise interferences.To solve the noise interference problem existing in ECG signals,in this study,the stationary wavelet transform combined with bivariate threshold method was used to denoise it.By performing an eight-layer stationary wavelet transform on the ECG signal,different wavelet coefficients will be obtained.The bivariate threshold function expression is used to process it,and the new wavelet coefficients will be obtained.Finally,the inverse stationary wavelet transform is used to realize wavelet reconstruction,and the ECG signal denoising is completed.The simulation results of Matlab proved that the proposed algorithm has higher accuracy and higher signal-to-noise ratio,and the signal-to-noise ratio reaches 84.5934 dB.Ⅱ.ECG signal waveform recognitionThe characteristic part reflecting the ECG signal is often the sudden change of the signal,so it is necessary to identify and detect the mutation point of the ECG signal.In this paper,quadric b-spline wavelet is used to make four-layer stationary wavelet transform for ECG signal after denoising.Waveform detection of R of ECG signal on the fourth scale.On the second scale,based on the correct detection of R waves,the peak and shape start and end points of Q and S waves are detected.In order to ensure the detection accuracy,the setting of the false detection and miss detection mechanism were also performed.Finally,P-wave,T-wave peak and start-stop point detection are realized at the five scale.The simulation results show that the proposed algorithm has high detection accuracy and the accuracy rate reaches 99.81%.Ⅲ.Classification of ECG signalsBy comparing the commonly used classification methods and the amount of data in the paper,the support vector machine classification model was selected for classification implementation.In the paper,to improve detection accuracy,the characteristics of time domain and wavelet domain was inputted,and the improved ant colony algorithm was used to optimize the parameters of SVM.The experimental results proved that the improved ant colony optimization SVM exhibited higher classification accuracy.The accuracy of the four classifications is as high as 98.89%.In summary,in the thesis,the SWT combined with the bivariate threshold is used to denoise the ECG signal,which not only overcomes the pseudo-Gibbs phenomenon but also guarantees the amplitude of the signal.On the basis of pretreatment,waveform recognition of ECG signal was carried out,and the detection of the characteristic parameters of each waveform of the ECG signal was realized.The four-class classification of ECG signals isachived by SVM and the MIT-BIH arrhythmia database was used to verify the classification algorithm.The classification algorithm has good classification performance.The research has certain practical significance,and it has certain promotion effect on the development of the ECG siginal automatic detection.