Machine Learning Classification of S-Band Microwave Scattering Measurements from the Forearm as a Novel Biometric Technique

Abstract

Biometrics use an individual's biological traits for personal identification. Various sensors have been used to obtain these measurements. Microwave biometric scans have recently gained traction as a non-contact technique due to their robustness to environmental lighting and unobtrusiveness. To evaluate microwave signature of human forearm as a biometric modality, an 8-antenna (Wi-Fi) data collection setup was developed and initially tested with foil-wrapped tubes of different geometric cross sections. The system was later evaluated by collecting microwave samples from human volunteers' forearms and classifying the data, from different antenna subsets, using Support Vector Machines and Naive Bayesian classifiers. Our results show that human identification via microwave signals is possible even with a subset of the above mentioned 8-antenna configuration.