Theoretical Study Of Multimode Biometric On DSP And ARM

With the popularization of various biometric identification technology and the widely application of personal identity authentication system,the single biometric,such as face,iris,vein and fingerprint recognition,will face a enormous difficulty.This paper is aim to combine the face and fingerprint with multi-feature and multi-biometric authentication system in an asynchronous serial mode,and proposed a multi-feature fusion algorithm in the embedded platform.In addition,designed a real-time moving target detection algorithm in order to reduce the power consumption and improve the performance of the systemAs for the DSP processor has the high powerful floating point computation ability and the ARM processor has the strongly capability to control.Combining with the advantages of DSP and ARM,a multimodal biometric authentication system is designed.In the DSP platform,the TI-RTOS system is combined with the RF5 software framework to manage the face image data acquisition,face detection algorithm,network communication,data transmission,serial port protocol and so on.ARM processor takes the role of achieving fingerprint image processing,image processing,finger identification and data communication.The data exchange and data synchronization of the system are realized through the UART and the network port.Firstly,we need to build a software development environment for the hardware platform.Combined with the computer vision library OpenCV,the Adaboost face detection algorithm based on DSP platform has been researched and the real-time detection performance has been tested.The face recognition algorithms such as Gabor,PCA,LDA and LBPs had been fully researched.A multi-feature fusion recognition algorithm has been studied and analyzed.It`s proved that the system is portable,easy to use,energy saving and so on.This paper consists of the following parts:1 ? Research on the related theories of image processing,machine learning algorithm,Gabor wavelet,computer vision and so on.2 ? Build DSP and ARM platform,initialization multi-modal biometric identification peripheral modules,including storage devices,fingerprint sensors,camera interface driver and so on.3?The application development of the multi task program based on the SYS/BIOS kernel,including the storage management and tasking multitasking procedures to achieve face detection algorithm.Design the communication protocol to achieve data exchange between DSP and ARM.4?In order to provide a higher level of security authentication,the multi-feature face recognition algorithm has been researched and realization in host computer.