Electrocardiogram beat classification using classifier fusion based on Decision Templates

This paper presents a ”Decision Templates” (DTs) approach to develop customized Electrocardiogram (ECG) beat classifier in an effort to further improve the performance of ECG classification. Taking advantage of the Un-decimated Wavelet Transform (UWT), which also serves as a tool for noise reduction, we extracted 10 ECG morphological, as well as one timing interval features. For classification we have used a number of diverse MLPs neural networks as the base classifiers that are trained by Back Propagation algorithm. Then we employed and compared different combination methods. Tested with MIT/BIH arrhythmia database, we observe significant performance enhancement using this approach.