Feasibility study on person identification utilizing frequency response functions of human fingers

Abstract

A novel biometric recognition method is presented for future personal devices, utilizing the frequency responses of the human body, especially the fingers, for personal identification. A series of experimental vibration modal analyses were conducted to measure the frequency response functions (FRFs) of individuals' fingers. In addition, a biodynamic lumped system model was constructed for the major components of a finger, including phalanges, joints, and skin. Analytical FRFs were then computed for a comparative study with the measured FRFs. In the person identification process, we employed an effective feature extraction method based on the correlation coefficient between frequency bins of the measured FRFs to extract the most effective set of frequency bins from the entire FRF spectrum. These extracted features were utilized to train a support vector machine for classifying individuals. In a controlled experimental setup, the classification results showed a maximum accuracy of 99%, affirming the feasibility of using the vibration responses of human fingers as a novel biometric method for person recognition.