Learning structural knowledge from the ECG

Abstract

We tackle the problem of discovering, without the “manual” aid of an expert, implicit relations and temporal constraints from a collection of dated events detected on temporally structured signals. The approach associates tightly signal processing and symbolic learning methods. It is illustrated on learning cardiac arrhythmias from ECGs.