A new ConvNet architecture for heartbeat classification

This paper presents a new convolutional neural network architecture for heartbeat classification. The architecture, that uses a reduced number of layers, with respect to other CNN used for heartbeat classification, is able to achieve high accuracy in heartbeat classification following the AAMI recommendations. In particular, using the well-known and researched electrocardiogram (ECG) MIT-BIH Arrhythmia database, the proposes convolutional neural network architecture shows similar performance when compared to the state of art classification algorithms using classical machine learning approaches.