ECG-based biometric recognition under exercise and rest situations

As a core technology in the field of information security, human biometric recognition has become the focus of researchers’ attention during the past few years, which is based on a myriad of biometric features including fingerprint, face, retina, etc. Due to the high difficulty of forgery, electrocardiogram (ECG) has a great potential to be applied into identification, while merely experiments on its rest situation have been worked on. In this manuscript, we overcome the oversimplification of previous researches, build our own ECG dataset containing signals under both exercise and rest situations, and evaluate the resulting performance on ECG human identification (ECGID), especially the influence of exercise on the whole experiment. By applying several established learning algorithms to our own ECG dataset, we find that current methods which can well support the identification of individual under rests, cannot equally present satisfying performance under exercise situations, therefore exposing the deficiency of existing ECG identification algorithms.