Drasko Nakic, Gjorgji Madzarov, Dejan Dordevic

Abstract: We want to introduce a lightweight system for arrhythmia recognition. The system is intended for outdoor use and is supposed to be implemented in mobile devices, primarily smart phones and smartwatches. We are only concerned about extracting pulses from an ECG stream and passing them to an Arrhythmia Recognition Module - ARM. These two basic tasks should be completed fast and effectively and give an alert when arrhythmia occurs. The ARM we designed relies on simple algorithm for feature extraction and gives an error of 2.007% with false/positives over false/negatives ratio of 19.