| The technology of automated analysis of electrocardiogram (ECG) is one of the most important methods for clinical examination, it can help doctors avoid recognizing the fussy images and improve the efficiency obviously. Also this technology has significance for the diagnosis of some diseases especially heart-blood vessel diseases.These days, automated analysis of ECG by computer is not widely used in clinical application, one reason is that the recognition of ECG is inaccurate, the other is that there is not accurate and uniform evaluating system including the diagnostic standard and testing database. So searching for new method is the basic approach for improving the effect of automated analysis of ECG and enlarging the application area.An algorithm of detecting ECG characteristic based on wavelet transform was presented in this paper. The characteristic features of ECG, including the peaks and limits of the individual QRS waves, P and T waves and the intervals of ST segment were detected exactly on multi-scales to improve the precision of detection. Radial Basis Function (RBF) neural networks combined with Bayesian regularisation was applied in arrhythmia classification. A method of classification is performed for four categories of beats based on RR-interval, and this method improves the detection accuracy of P waves in pathologic ECG waves. In this thesis, the extractions of ECG features were simulated in Matlab platform which was used as calculation engine also. The algorithm was evaluated on several manually annotated databases, such as MIT-BEH Arrhythmia, and European ST-T databases. The QRS detector obtained an ideal effect. So all conclusion proved that this algorithm was reliable.Further research and plenty of experiment were carried through in this paper aiming at improving the precision of detection of ECG and establishing valuable diagnostic standard. The technology of automated analysis of ECG was still a huge task, which had a prospective future. |
PDF Full Text Request