ECG-Based Biometric Key Generation Using Principal Component Analysis and Machine Learning

Biometric authentication techniques are fast becoming imperative methods for secure identifications in a wide range of applications, while most of the traditional systems are easily spoofed or forged. In this paper a new approach of deriving a unique biometric key from an electrocardiogram signal is presented owing to physiological uniqueness of heart activity. In this respect, by focusing on RR intervals extracted from ECG signals, the PCA is applied in order to reduce its dimensionality and then form a compact and distinctive biometric key. Thereafter, a Random Forest classifier was used in evaluating the effectiveness of features, where a high classification accuracy of 98.5% is obtained. The uniqueness, stability, and security of the biometric key generation process were good, while the false accept rate was 0.01% and the false reject rate was 0.02%. The findings further support that ECG signals, more precisely RR intervals, form an efficient source of biometric information in secure identification. This observation states that approaches like these will face promising real-world applications in health care and continuous authentication systems.