基于HMM的嵌入式人脸识别系统研究

Research of an Embedded System for Face Recognition Based on HMM

系统选用S3C2410A处理器作为硬件平台,Linux操作系统作为软件平台,采用OV7640+OV511的结构实现图像信息采集,使用基于V4L的方法编写图像处理、检测和识别程序,实现了快速、准确识别人脸的功能;并对相关浮点算法进行了改进,极大地提高了ARM处理器上图像预处理的速度。采用基于隐马尔可夫模型(HMM)的方法,简化了人脸检测和识别的过程。

The system uses a CMOS digital imaging sensor OV7640 plus OV511, the S3C2410A processor and the Linux operation system for the image processing. The program is written by the method of V4L, including image-pretreatment, face detecting and recognition. The float image-pretreatment algorithm is improved to enable highly image-pretreatment implementation for ARM9 microprocessor. This new design is suitable for face detection and recognition, thus to make an important step towards low-cost and portable systems. The method based on Hidden Markov Model predigests the process of face detection and recognition.