Face Recognition System Based DSP

Face recognition system gets more and more attention because of its tremendous application prospects for security authentication systems, credit card verification, file management, video conferencing, human-computer interaction, security system (criminal identification) etc. The traditional face recognition system relys on the support of computer hardware and operating system that bears limitation like poor real-time, high system costs and and large volume, and it is difficult to meet the needs of the general consumer market. The face recognition technology which based on digital signal processor (DSP) can have small size, low cost and high reliability by combining with embedded system design concepts. As a result, the research and development of face recognition technology based on DSP has become a research hot spot in the field of pattern recognition and artificial intelligence.This paper leads to the needs of researching the based-DSP face recognition system, based on introducing the research background and main methods of the face detection and recognition technology. The face detection algorithm based on skin color and the Gabor + PCA face recognition algorithm are firstly introduced in detail. The design process of embedded one Face recognition system is respectively described from system structure design, hardware and low-level driver design (the main chip design, driver design), system control and the implementation of algorithm with software. The system structure and working mode is designed, and paper gives the technical requirements of embedded one Face recognition system. Paper introduces TI TMS320DM642 chip, FLASH, and SDRAM memory chips, and achieves the design of power supply , memory module, serial and video input and output, the DSP physical level driver is implemented with the CSL library as well. The part of the software design introduces the system main control process, the transplantation of face detection and recognition algorithms, and how to optimize the algorithm by software optimization. Finally paper achieves a based-DSP face recognition system, and lists out the steps and method of debugging on the DM642 hardware development board.The experiments show that the face recognition system based on DM642 hardware platform, with the algorithms proposed on the paper, achieves the real-time detection of front face in the case of a single human face, the rate of detection is more than 90%, and rate of face recognition is more than 75%. The results indicate that DM642 chip is capable to deal with real-time video processing, and the algorithms used on the paper achieve higher rate of recognition and detection for frontal face, the system has great utility.