Abstract
Heart disease is the third of the top ten causes of deaths in Taiwan these years. The heart disease cause of the cardiovascular disease is unexpected hence developing a real-time heart disease detector system is important. In this thesis we implement an ECG signal classification system with these stages: signal pre-processing, wavelet transform, and signal classification. In the first stage, median filter is used to reduce the noise and baseline wander on the ECG signal. Then we detect the QRS complex and extract the features of QRS complex by wavelet transform. Finally, we use nearest neighbor method to classify each pattern of QRS complex. A QRS complex is classified into one of the seven types: normal beat (N), left bundle branch block beat (L), right bundle branch block beat (R), premature ventricular contraction (V), atrial premature beat (A), ventricular escape wave (E), and ventricular flatter wave (I). The QRS complex classification is tested on two databases: MIT-BIH Arrhythmia Database and the European ST-T Database. The experimental results can be divided into two parts: QRS complex detection and ECG signal classification. In QRS complex detection, the accurate detection rates are above 99% in these two databases. The recognition rates are about 97% and 99% in the MIT-BIH Arrhythmia Database and the European ST-T Database, respectively.