首页> 外文会议>2014 12th International Conference on Signal Processing >A low-cost fully pipelined architecture for fingerprint matching
【24h】

A low-cost fully pipelined architecture for fingerprint matching

机译:用于指纹匹配的低成本全流水线架构

获取原文

摘要

Fingerprint matching is a key procedure in fingerprint identification applications. The fingerprint-matching algorithm based on minutiae is one of the most typical algorithms that can achieve a reasonably correct recognition rate. Performance and cost are two critical factors when implementing minutia-based matching algorithms in most embedded applications. A low-cost, fully pipelined architecture for minutia-based fingerprint matching is proposed in this paper. A regular matching unit with a pipeline of 13 stages is designed as the core of the architecture, interfacing with a two-port RAM and a DDR3 controller. We implemented the whole architecture on a Xilinx FPGA board with the Virtex VII XC7VX485T chip. The matching unit can run with a frequency of 330 MHz on the chip, which leads the system to achieve a throughput of about 430000 fingerprints per second when processing typical datasets. The unit only occupies 568 slices, which is less than 1% of the available chip resources. The board only consumes 16 W of power when run. The architecture can gain about twice the throughput of the 2.93 GHz Intel Xeon5670 CPU at a low logic cost and power.
机译:指纹匹配是指纹识别应用程序中的关键过程。基于细节的指纹匹配算法是可以实现合理正确识别率的最典型算法之一。在大多数嵌入式应用中实施基于细节的匹配算法时,性能和成本是两个关键因素。本文提出了一种用于基于细节的指纹匹配的低成本,全流水线架构。具有13级流水线的常规匹配单元被设计为体系结构的核心,并与两端口RAM和DDR3控制器接口。我们使用Virtex VII XC7VX485T芯片在Xilinx FPGA板上实现了整个架构。匹配单元可以在芯片上以330 MHz的频率运行,这使得系统在处理典型数据集时每秒可实现约430000个指纹的吞吐量。该单元仅占用568个切片,不到可用芯片资源的1%。该板在运行时仅消耗16 W的功率。该架构可以以低逻辑成本和低功耗获得2.93 GHz Intel Xeon5670 CPU的大约两倍的吞吐量。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号