刘培龙, 黄乐天, 林凌鹏. 基于FPGA的神经元相关性分析的设计与实现[J]. 微电子学与计算机, 2012, 29(5): 119-123.
引用本文: 刘培龙, 黄乐天, 林凌鹏. 基于FPGA的神经元相关性分析的设计与实现[J]. 微电子学与计算机, 2012, 29(5): 119-123.
LIU Pei-long, HUANG Le-tian, LIN Ling-peng. Design and Implementation of Neural Correlation Analysis Based on FPGA[J]. Microelectronics & Computer, 2012, 29(5): 119-123.
Citation: LIU Pei-long, HUANG Le-tian, LIN Ling-peng. Design and Implementation of Neural Correlation Analysis Based on FPGA[J]. Microelectronics & Computer, 2012, 29(5): 119-123.

基于FPGA的神经元相关性分析的设计与实现

Design and Implementation of Neural Correlation Analysis Based on FPGA

  • 摘要: 利用Verilog HDL语言,Xilinx的ISE平台实现了神经元相关性分析的设计.首先对神经元相关性分析的理论和软件实现的方法进行了简单介绍,然后对相关性分析的主要模块进行了设计,最后用ModelSim进行了功能仿真和时序仿真,用ISE做了逻辑综合与实现以及性能分析.所选FPGA器件xc5vlx220-2ff1760逻辑资源消耗只占7%,最高时钟频率可以达到240Mhz左右.只需要48个时钟周期就可以实现两个神经元之间相关性的计算,也就是200ns.64通道的情况下需要0.4ms,而用软件实现的方法至少需要几秒的时间,这样可以对神经元之间的相关性进行实时性分析.

     

    Abstract: Implementing the design of neural correlation analysis using Verilog HDL and Xilinx ISE platform. Firstly, introduce the theory and software implementation method of neural correlation analysis. And then design the main module of the correlation analysis. Finally, do the functional simulation and timing simulation using ModelSim, synthesize and implement the design based on ISE platform. Only 7% logic resources are consumed of the FPGA device xcSvlx220-2ff1760. The maximum frequency can reach 240MHz. It only needs 48 clocks to implement the correlation analysis between two neurons that is 200ns. 64 channels need 4ms. However, if use the software method, it needs some seconds at least. So, it can realize the real time neural correlation analysis.

     

/

返回文章
返回