Open AccessAn Identity Authentication Method Based on Multi-modal Feature Fusion
Author(s) -
Jiabo Chen,
Linxi Cai,
Yifeng Tu,
Rui Dong,
Diankun An,
Baili Zhang
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1883/1/012060
Subject(s) - computer science , modal , speech recognition , pronunciation , liveness , biometrics , facial recognition system , classifier (uml) , identity (music) , authentication (law) , artificial intelligence , reading (process) , pattern recognition (psychology) , computer security , linguistics , philosophy , physics , acoustics , polymer chemistry , chemistry , programming language
At present, many judicial organs have adopted the daily attendance system based on face recognition to strengthen the supervision of community correction personnels. In order to prevent a few personnels from using pre-prepared photos and videos to deceive the face recognition system, this paper proposes an identity verification scheme with liveness detection based on dynamic combination of multimodal features. The main idea is as follows. Firstly, during face verification, the user is required to read out random numbers on the screen. Secondly, generating dynamic combination of speech, voiceprint, lip-reading and other verification methods according to the user’s risk personas, so as to achieve a balance between convenience and security. In addition, in view of the low accuracy of lip-reading recognition in practice, this paper changes the traditional lip-reading recognition based on morpheme to lip-reading recognition based on classifier. By optimizing the interactive content, the distinction of pronunciation between different Chinese characters is increased, and the accuracy of lip-reading recognition is significantly improved.