Study Of Individual Cardiogram Waveform Automatic Selection In loiter

The paper studied the automatic selection method on dynamic electrocardiogram (Holter). Firstly, the features of electrocardiogram (ECG) waveform were extracted by wavelet transform. Secondly, clustering working was done on first 3000 ECG waveforms by self organization map neural network (SOM), from which, labeled sample waveforms were gotten. In the end, back propagation (BP) neural network were used for ECG waveform classification. From the experiment result, the ECG R wave recognizing rate was up to 99.5% by wavelet transform. According to the labeled sample that clustered by SOM, BP neural network can correctly classify ECG wave in 95%. The methods for automatic selection of Holter data can be used for real work.