SIGNAL VALIDATION FOR CARDIAC BIOMETRICS

Medical biometrics offer direct solutions to the liveness and impersonation detection risks which dominate traditional biometric modalities, like the iris, the face or the fingerprint. The electrocardiogram (ECG) is a cardiac signal which falls under this category, and which has lately drawn interest from the biometrics community. This paper presents a novel recognition method based on ECG signals, which enhances the AC/LDA feature extraction algorithm, by incorporating the periodicity transform (PT). It is demonstrated that PT is a powerful tool not only in assessing the matching validity of the signal, but also in handling heart rate changes. The performance of the system over 52 subjects is 92.3%.