Person identification using electrocardiogram and deep long short term memory

There have been numerous attempts in recent years to build trustworthy biometric identification systems. One of the newest approaches for human identification is the electrocardiogram (ECG) biometric, which not only provides distinctive characteristics of individuals but also has a very low chance of being faked. In this study, a person identification system based on ECG is proposed, where the differentiated ECG and its statistical measurements are used as features sets and for the classification of ECGs, long-short term memory (LSTM) is considered. The proposed method has been found to be more accurate than state-of-the-art methods based on results from a variety of normal and abnormal pathological ECG datasets.