Research And Development Of Face Recognition And Authentication System For Mobile Phone

Biometric authentication technologies received increased attention recently in academic research and practical applications. Compared with other biometric technologies, face recognition provides a more natural and easier approach for human identification and has great theoretic and practical values.Mobile phone is the main agent of the mobile communication and now an accepted part of everyday life. This provides a wide range of potential applications for mobile servers. But in the most of systems, the only security to prevent unauthorized use of the mobile phone is the digit password. This is neither satisfactory nor reliable enough. To addressing these questions, this thesis studies face recognition and develops a face recognition and authentication system for mobile phone. The main contests of the thesis are listed as followings:1. Based on the requirement and functional analysis of the system, the workflow and functional modules of the system is made sure. Then based on the functional module, the face authentication system's scheme of the interface, enrollment, verification, settings are designed and the development platform and the evaluation criterion are presented.2. Image preprocessing technologies for face recognition are studied. The geometric normalization technique and the photometric normalization technique for system are adopted. And the image preprocessing technology is optimized and tested on the mobile phone. The experimental results show the image preprocessing enhances the performance of face recognition.3. Advantages and disadvantages of current face recognition technologies are discussed. Fisherface is adopted as the system's algorithm due to the limited amount of resource available in the mobile phone. And Gabor wavelet is selected to enhancing the robust to lighting variations. The principle of the adopted face recognition algorithm is introduced detailedly. Functions of the algorithm are tested using the adopted evaluation criterion. The experimental results indicate that the adopted algorithm provides better performance compared with Fisherface algorithm. And the face recognition algorithm is optimized and tested on mobile phone. The experimental results show the optimized algorithm is faster and more accuracy than the non-optimized algorithm.4. The face recognition arithmetic is improved for the mobile phone's CPU speed, memory size and the lack of floating point units limitations. Based this a real-life face recognition and authentication system is developed on mobile phone and it can achieve automatically and Real-timely the functions of face detected, eyes located, feature extracted, face image preprocessing, user enrollment, user verification and threshold settings. The performance of the system is tested in public face database and real-life application environment. The experiment results show that the system approaches achieved high recognition rate (ERR >90%) and real-time requirement.