ECG Signal Classification using Deep Neural Networks with Ensemble Techniques

Electrocardiogram (ECG) is a vital signal, used to identify cardiovascular diseases like arrhythmias, heart attacks and coronary heart diseases. The classification of ECG signals help doctors to give correct treatment of a particular disease. Our proposed work based on classification of five classes. This research study has used a stack of neural networks, ensemble techniques like Dropout, global maxpooling and data sets like MITBE PTBDB which are considered as the open source databases. Data pre-processing is done by using resampling and normalization techniques. The proposed CNN-BiLS TM-Attention mechanism has obtained an accuracy of 98.9% and F1score of 0.947. Comparing to existing methods, the proposed method improves the accuracy.