Research And Implementation On Embedded Biometrics System Fusing Face And Fingerprint Recognition

Biometrics is used more and more widely in the field of authentication for its security and convenience. It can solve the problems of traditional identification. There are many biometrics, such as fingerprint, face, palm prints, iris, retina, voice, and gait. Fingerprint and face are the two most successful ones of all because of their high universality, unique, permanent, availability, and acceptability. In recent years, research of fingerprint identification and face recognition has made significant progress, and the speed and rate of identification are greatly improved. But one single biometrics neither fingerprint nor face can achieve perfect performance. For example, somebody has no fingerprint, and twins can’t be distinguished by face recognition. Consequently, fusion of multiple biometrics becomes the trend with the development of the research.Recently, embedded technology is developing rapidly. A wide range of embedded products play an important role in industrial control, defense, security, and communications. It becomes more and more evident that the performance is higher and higher, the cost is lower and lower, and the deployment is faster and faster. But some embedded biometrics products in the market have some problems. For example, someone without fingerprint can’t use the fingerprint identification products, and the face recognition products haven’t got enough accurate identification rate. Therefore, the combination of biometrics and embedded technology can provide portable and high performance products. In this thesis, an embedded multi-biometrics system is designed and implemented.Firstly, a multi-biometrics solution is involved. It is based on fingerprint identification and face recognition is proposed, which solves the problems of single biometrics. A novel fusion method based on confidence is proposed, which can make sure the high identification rate and good user experience.Secondly, a complete system of practical system design, hardware design and software design is introduced. The principle of the system design has guiding value for electronic product design. The hardware design scheme of ARM+DSP is adopted. ARM is used for system control and human-computer interaction, and DSP is used in the operation of the biometric identification algorithm, which can meet the requirements of the algorithm to a hardware computing, at the same time reducing development difficulty, and promoting the efficiency of development. The software design addresses in detail about the process of the embedded software development, the choice of the embedded operating system, and the development and deployment of the application.Finally, Key technology and solution are introduced. The fingerprint stitching algorithm is proposed. The two level search strategies not only ensure the accuracy of the algorithm, but also guarantee the real-time of stitching. Then the algorithm of fingerprint segmentation is involved. The result of experiment shows that the algorithm has the very good adaptability to different fingerprint sensors. In the end, the dual-CPU solution is proposed in detail.