Hardware Implementation of Neural Networks Based on DSP48 Slice of Virtex-4 FPGA

摘要：

It is well known that a perceptron, a type of artificial neural networks, has a lot of application in the area of classification.pattern recognition, clustering, among others. This paper presents a hardware implementation approach for multilayer perceptron neural networks using only one DSP48 slice and two block RAMs in Xilinx Virtex-4 SX35 FPGA (Field Programmable Gate Arrays). As a typical instance of perceptron neural network application, we have implemented a handwritten digit recognition system on our hardware implementation. Experimental results show that our hardware approach is simple, flexible, low cost and high performance, which provides a trade-off between resource requirements and speed.

关键词： Perceptron Neural Networks FPGA DSP48 slice Block RAM

作者: Jiajin Zhang Lichang Chen

作者单位: Department of Computer Science and Technology School of Information Engineering Yunnan Agriculture U Department of Computer Science and Technology School of Information Engineering Yunnan Agriculture U

会议类型: 国际会议

会议名称: 2011 3rd International Conference on Computer Engineering and Applications(2011第三届计算机工程与应用国际会议 ICCEA2011)

会议地点: 海口

会议语种:英文

页码: 257-261

在线出版日期: 2011-07-15（万方平台首次上网日期，不代表论文的发表时间）