Biometric efficient identification algorithm and network authentication system with robust liveness check
Jiaqi Zhu · 2010
datasets show that our methodology is much more efficient than many of the wellknown methods.We will introduce our multimodal biometric authentication system that integrates speech recognition, speaker verification, face verification and lip shape estimation together to provide a user-friendly, accurate and robust authentication.These biometric features have been chosen on the basis of their usability in the public domain, maturity of technology and minimal inconvenience to the user.New algorithms for improving the performance of speaker verification are presented.The idea is to assign to every enrolled user a set of user-specific parameters.Novel liveness check mechanisms are introduced and they are the audio liveness check and the joint audio-video liveness check.The audio liveness check uses speech recognition.The joint audio-video liveness check uses lip shape estimation with speech recognition.Fusion and decision making algorithms are also presented.Challenge response database and systems robustness are discussed.Methods of estimating the system's robustness are discussed.Experiments are conducted to show the system's robustness against replay and brute-force attacks.