Biometric identification with high frequency electrocardiogram: Unregistered user refusal method and performance evaluation

As a new modality for biometric identification, electrocardiogram-based identification technique has been developed. We proposed a technique with high frequency component of electrocardiogram (HFECG) in QRS segment. In this report, an unregistered user refusal algorithm was combined with the artificial neural network based waveform classifier. The refusal function was realized by simple thresholding technique. HFECGs from twenty collaborators were used for supervised learning. Twenty HFECGs from the same collaborators were tested and false acceptance rate (FAR) and false rejection rate (FRR) were evaluated. Ten HFECGs from other collaborators were also tested to find unregistered user refusal performance. The results show that FAR and FRR in the registrants can be kept within 1%, however, unregistered user refusal performance was not acceptable under the same condition.