软硬件协同设计方案与经验性PCA算法的FPGA实现

Hardware and Software Co-design and Empirical PCA Algorithm on FPGA

在安全监控领域,人脸识别系统的需求日益增长,这要求系统能够从摄像头采集、人脸识别到结果显示均可靠实现.文中提供一种软硬件协同设计方案.首先通过软件设计有效地选择比特位宽、定点数算术用于指导硬件设计.其次,通过软件实现数据的采集、显示、传输和系统控制,FPGA实现人脸识别算法提高算法速度和降低系统复杂度.阐述了软硬件协同方案和PCA算法的FPGA实现.相比传统的纯软件实现人脸识别算法,提高了效率,识别率74％,错拒率4％,误识率6％.最后,研究了PCA算法相关门限阀值的设置,发现通过合理的范围预判和门限阀值设置可以有效地提高4％～6％的识别率.

In ecurity surveillance, there are growing demands for recognition system, which require the system to collect data from the camera, face recognition to achieve reliable results, display screen to show the right images.This article provides a hardware and software co-design solutions.First of all, through software design , we effectively choose the perfect bit wide and right fixed-point arithmetic to guide hardware design.Secondly, we use software to achieve data collection, display, transmission and system control, use FPGA to achieve face recognition algorithms to improve the algorithm speed and reduce system complexity.This paper explicates the scheme of the hardware and software co-design and implementation of PCA algorithm on FPGA.Compared with the traditional face recognition algorithm only software, this greatly improves efficiencyThe recognition rate is 74％, the false rejection rate is 4％, and the error rate 6％.Finally, we study the threshold setting of data which is relevant to PCA algorithm, we find that a reasonable Range judgment and threshold setting can efficiently Improve the recognition rate of 4％ to 6％.