The Implementation of Feedforward Backpropagation Algorithm for Digit Handwritten Recognition in a Xilinx Spartan-3

Panca Mudji Raharjo, Mochammad Rif'an, Nanang Sulistyanto

Abstract


This research is aimed to implement feedforward backpropagation algorithm for digit handwritten recognition in an FPGA, Xilinx Spartan 3. This research is expected to give a contribution such as the feedforward algorithm design in VLSI technology based on FPGA, the practice module of Xilinx Spartan-3 development board and further research in artificial neural network and FPGA field in Electronics Laboratory.

The feedforward backpropagation algorithm is used to recognize 10 objects. The feedforward backpropagation network consists of two layers, 36 input unit which is the feature vector of object, 10 hidden neurons, and 10 output unit. The first layer activation function is tansig and second layer activation function is purelin.

The multipliers use 18 bits. The proposed design fits into the smallest Xilinx FPGAs3.

Index Terms—feedforward backpropagation network, digit handwritten recognition, FPGA, Spartan-3.


Full Text:

PDF

Refbacks

  • There are currently no refbacks.