Local-Global Attention Enhanced MobileNet for ECG Arrhythmia Classification

As a kind of common cardiovascular disease characterized by abnormal rhythm of cardiac electrical activity, the disordered electrical signal conduction may lead to serious consequences such as cardiac dysfunction, stroke and even sudden cardiac death. In view of clinical diagnosis, this article puts forward the combination of local-global attention mechanism lightweight classification model, and noise reduction preprocessing using improved wavelet thresholding, focusing on normal stroke, left bundle branch block, right bundle branch block, atrial premature beat and ventricular premature beat five kinds of typical arrhythmia classification study.