Research On Classification Of Arrhythmia Based On Wavelet Transform And Neural Network

As the incidence and mortality of heart disease increase year by year,heart disease has seriously threatened human health,thus long-term cardiac monitoring is crucial for heart disease patients.Arrhythmia is much common in heart disease.It is also indispensable for the detection of the Arrhythmia.Electrocardiogram analysis(ECG)is one of the most common detection methods in the clinic.Doctors need to manually detect the type of heart disease in a large amount of monitoring data,which consumes much time and medical resources.Moreover,the current automatic Arrhythmia classification algorithm has problems that the features do not have physiological significance,the sample is small and the data cannot be fully utilize.In response to these problems,this thesis is to use more sample of ECG signal to complete the denoising,feature extraction,automatic classification of five Arrhythmia.The main work of the thesis are as follows:1.Since some features do not have physiological significance and can not assist doctors in making diagnosis,this thesis proposes an adaptive Discrete Wavelet Transform to extract the features that have clinical significace to form the feature matrix,so as to make full use of data information,and abtain higher classification accuracy.2.To solve the problem of high dimension of feature matrix of the comprehensive feature extraction and the large sample,this thesis proposes the Principal Component Analysis to reduce the dimension of the feature,convert the original high-dimensional feature matrix into a low-dimensional feature matrix,and remove redundant information,so as to reduce the amount of calculation and speed up the operation.3.Aiming at the low classification accuracy of traditional Artificial Neural Network and the slow convergence speed of error function,this thesis proposes an improved Levenberg-Marquardt algorithm based on traditional Artificial Neural Network classification,can improve classification accuracy obviously,and can recognise five Arrhythmia under the MIT-BIH Arrhythmia Database,namely Normal sinus rhythm,Left Bundle Branch Block,Right Bundle Branch Block,Premature Ventricular Contraction,Premature Atrial Contraction,and can achieve 98% accuracy,at the same time,accelerate the convergence speed of the error function.