Human asymmetry in multibiometric recognition

Abstract

This study examines human fluctuating asymmetry as a tool to improve multibiometrics. A new technique called Bilateral Processing (BP) was implemented to analyze discordances in left and right trait sides. BP tested visible and infrared spectrum trait images using Cross-Correlation, Wavelets and Neural Networks. Selected traits were tooth, ear, iris, fingerprint, nose and cheek. Acoustic BP was also implemented for skin vibration asymmetry evaluation during voiced sounds and compared to MFCC plus Vector Quantization speaker recognition system. Image and acoustic BP investigated bilateral sides of 9 adult male brothers during one year. Results achieved 100% identification in all biometrics treated with BP, compared to a maximum 44% of correct identification without BP. Apart from ameliorating performance, the study concludes that body structure analyses incorporating biognostic mechanisms can complement traditional recognition approaches.