计算机应用 ›› 2010, Vol. 30 ›› Issue (3): 854-856.

• 典型应用 • 上一篇    

基于DSP的人脸识别算法实现与优化

邹垚1,张超2   

  1. 1. 上海理工大学光电信息与计算机工程学院
    2. 上海理工大学南校区光电信息与计算机工程学院
  • 收稿日期:2009-07-29 修回日期:2009-11-13 发布日期:2010-03-14 出版日期:2010-03-01
  • 通讯作者: 邹垚
  • 基金资助:
    上海市研究生创新基金

Implementation and optimization of face recognition algorithm based on DSP

  • Received:2009-07-29 Revised:2009-11-13 Online:2010-03-14 Published:2010-03-01

摘要: 为了在嵌入式系统中实现实时视频图像人脸识别,提出了基于TI系列TMS320DM642的快速人脸检测系统设计方案,包括系统软件和硬件优化方案。首先选择实用有效特征,简要介绍人脸检测和人脸识别算法的基本原理,针对该算法实现原理详细阐述硬件系统设计方案,介绍各单元结构和原理。最后,在设计的硬件系统上进行算法移植,详细介绍了系统的优化方案,实现了嵌入式快速人脸识别系统的研制。通过分析测试结果,系统可靠运行,优化后系统运行速度提高,能够实现实时视频图片人脸识别。

关键词: TMS320DM642, 人脸识别, Adaboost算法, Gabor变换, 优化

Abstract: In order to realize the real-time face recognition in video by embedded system, the design scheme of fast face recognition system based on TI TMS320DM642 was presented, and optimization policy was proposed in terms of both hardware and software. Firstly, effective features were selected, and the algorithm of face detection and recognition was briefly introduced, then according to the algorithm, the design scheme and function units of hardware system were described. Finally, the algorithm was transplanted onto the hardware system, then the optimization policy was applied, and fast embedded face recognition system was implemented. The experimental results demonstrate the system works reliably, the running speed of algorithm was improved markedly, and real-time face recognition in video was realized.

Key words: TMS320DM642, face recognition, Adaboost algorithm, Gabor transformation, optimization