Deep Learning Algorithm for the People Identification Using their ECG Signals as a Biometric Parameter

Reliable identification and authentication methods are a growing concern in various domains. Despite facial recognition, fingerprint and iris recognition being common options in the field of biometrics, their inherent vulnerabilities have led to exploration of new options. In particular, electrocardiogram (ECG) signals are gaining interest as a promising biometric feature due to its unique and difficult-to-falsify characteristic. In addition, the ECG is a noninvasive signal that can be collected quickly and easily. Therefore, in this paper, we propose a deep learning algorithm capable of performing biometric identification using ECG signals, highlighting the potential of these signals as a unique biometric feature. Our work focuses on utilizing a Transformer neural network for biometric identification based on pattern recognition of ECG signals. This type of network includes attention layers to highlight the most relevant parts of the signal, and we describe the process of training and evaluating the model.