The Cardiogram Signal Analysis And Realization Based On Wavelet Packet Transform

Cardiac diseases are one of major diseases that endanger human's health. Analysis of ECG(Electrocardiogram) is an important means of clinical diagnosis of cardiac diseases. And it has been a research hotspot for many years at home and abroad that use computer to analyze ECG and diagnose cardiac diseases.First of all, this study discusses the basic principles of the two filtering methods: wavelet analysis filter and wavelet packet analysis filter in ECG pre-processing, evaluating the two comparative denoising effect and giving the results of the evaluation.Secondly, this article studies in the foundation of predecessor and uses the method which the wavelet packet transformation(WPT) and the derivative unify to carry out the ECG signal feature extraction, according to the characteristics of each band using different positioning methods. Use the wavelet packet transformation to locate the R wave first, then according to the relations between R wave with other waves to determine the beginning and end spot of Q,S wave on the basis of the R wave accurate localization. Finally, according to the signal energy in each band of the probability distribution of sub-space this study use the characteristic of part of the decomposition tree node coefficient of the relative wavelet packet energy signal covers the main energy information to extract these eigenvector.Finally, this study classifies four types of ECG signal in the MIT / BIH database,which are: normal cardiac arrest, with a large number of ventricular premature beat, with serious baseline drift and with the pacemaker based on wavelet packet energy eigenvector. Using of method of WPT combined with BP NN and WPTcombined with SVM to evaluate the four types of typical ECG signal. According to the training process and error performance curve of these two methods are to give the classification of evaluation results.