Plantar Biometrics for Edge Computing

Abstract

Biometric systems are getting integrated into our daily life as the needs for authentication are increased rapidly. In smartphones fingerprint and face identification are used already widely as a method for user authentication. A relatively novel area of biometrics is the usage of plantar biometrics, foot sole features, to verify human identities. There are several approaches to utilise plantar biometrics but most of the proposed approaches require bulky, obtrusive or an immobile design. In this paper, we introduce a unobtrusive biometric system based on a shoe wearable, which is able to authenticate individuals with the assistance of Neural Network Classifier. The implemented system is evaluated on $10$ individuals achieving $94.3\%$ accuracy with a loss of $1.87$. Furthermore, the learning and authentication part takes place on the edge which has numerous benefits towards the performance but also the security aspects of the system.