Model-based, data-driven and hybrid biometric person authentication using cardiac signals of electrical, mechanical...

The article is devoted to development, investigation, and comparative analysis of the new model-based, data-driven and hybrid authentication technologies utilizing cardiac signals of different physical nature (electrocardiogram (ECG), seismocardiogram (SCG) and photoplethysmogram (PPG)). The proposed model-based and hybrid pipelines rely on representing cardiac signals as cyclic random processes with variable rhythm and on rhythm-adaptive processing for cycle alignment and feature extraction. The rhythm-adaptive estimate of the mathematical expectation is used as a compact subject-specific representation of cardiac-cycle morphology. Experiments were conducted on ECG and SCG signals from the CEBS database and PPG signals from the PTTPD database. For each modality, model-based, data-driven, and hybrid authentication technologies were evaluated using classification metrics and operating-point biometric verification characteristics, including False Rejection Rate (FRR), False Acceptance Rate (FAR), and Half Total Error Rate (HTER). The results show that hybrid technologies generally provide the best trade-off between accuracy, verification reliability, interpretability, and computational cost. Under the within-session evaluation protocol, ECG- and SCG-based hybrid authentication reached the highest average accuracy values, while PPG-based authentication showed slightly lower but still high performance. The study demonstrates that incorporating a mathematical model of cardiac signals and rhythm-adaptive processing into authentication pipelines can substantially improve biometric performance compared with purely data-driven baselines.