Biometric authentication utilizing vibrational characteristics of human fingers

Abstract

Biometric authentication is an emerging technology in mobile devices and various industry applications. Existing image-based methods such as fingerprint, face and iris recognition have developed but they have inherent disadvantages in terms of sensor miniaturization and user accessibility. This research suggests a new biometric authentication process that uses the vibration characteristics of human body, especially fingers. The frequency response function (FRF) were used to distinguish individuals. The concept of vibration modal analysis was conducted to measure FRFs of several fingers and its resonance frequencies were observed pursuing to extract robust and repetitive vibration characteristics of fingers. Data-driven feature extraction and model-based feature extraction were applied to the measured FRFs to extract various types of features. In particular, model-based feature extraction model the major tissues of finger such as phalanges, joints and skin in biodynamic system to calculate the analytic FRFs. The mass, spring, damping coefficient parameters were estimated by comparing them with experimental FRFs. The selected feature vectors were trained on the support vector machine (SVM) and for the classification of individuals. In addition, appropriate signal processing methods and long-term data accuracy were tested through the trained classifier. The classification results using the magnitude of FRFs showed 99% accuracy at maximum in a controlled experimental setup.