Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA

A hardware implementation based on Field Programmable Gate Array (FPGA) of a single-layer feedforward neural network for handwritten digit recognition has been developed. The effect of the network coefficient bit depth on the recognition accuracy and FPGA hardware costs has been studied. The neural...

Full description

Saved in:
Bibliographic Details
Main Authors: E. A. Krivalсevich, M. I. Vashkevich
Format: Article
Language:Russian
Published: Educational institution «Belarusian State University of Informatics and Radioelectronics» 2025-04-01
Series:Doklady Belorusskogo gosudarstvennogo universiteta informatiki i radioèlektroniki
Subjects:
Online Access:https://doklady.bsuir.by/jour/article/view/4117
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849340610762244096
author E. A. Krivalсevich
M. I. Vashkevich
author_facet E. A. Krivalсevich
M. I. Vashkevich
author_sort E. A. Krivalсevich
collection DOAJ
description A hardware implementation based on Field Programmable Gate Array (FPGA) of a single-layer feedforward neural network for handwritten digit recognition has been developed. The effect of the network coefficient bit depth on the recognition accuracy and FPGA hardware costs has been studied. The neural network was trained using the MNIST handwritten digit database. The neural network prototype was implemented as an IP core on the ZYBO Z7 debug board. The developed prototype was used to perform experiments with different bit depths of neural network coefficient representation. Graphs of recognition accuracy and the amount of FPGA hardware resources depending on the bit depth of neural network coefficient representation have been constructed. The coefficients obtained as a result of neural network training have been analyzed using decomposition into bit planes. It has been shown that 5 bits are sufficient to represent neural network coefficients, since they contain the main information learned by the network, ensuring economical use of FPGA resources and high recognition accuracy (92.4 %).
format Article
id doaj-art-878aa794e9d340219070fbc436543dcb
institution Kabale University
issn 1729-7648
language Russian
publishDate 2025-04-01
publisher Educational institution «Belarusian State University of Informatics and Radioelectronics»
record_format Article
series Doklady Belorusskogo gosudarstvennogo universiteta informatiki i radioèlektroniki
spelling doaj-art-878aa794e9d340219070fbc436543dcb2025-08-20T03:43:52ZrusEducational institution «Belarusian State University of Informatics and Radioelectronics»Doklady Belorusskogo gosudarstvennogo universiteta informatiki i radioèlektroniki1729-76482025-04-0123210110810.35596/1729-7648-2025-23-2-101-1082067Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGAE. A. Krivalсevich0M. I. Vashkevich1Belarusian State University of Informatics and RadioelectronicsBelarusian State University of Informatics and RadioelectronicsA hardware implementation based on Field Programmable Gate Array (FPGA) of a single-layer feedforward neural network for handwritten digit recognition has been developed. The effect of the network coefficient bit depth on the recognition accuracy and FPGA hardware costs has been studied. The neural network was trained using the MNIST handwritten digit database. The neural network prototype was implemented as an IP core on the ZYBO Z7 debug board. The developed prototype was used to perform experiments with different bit depths of neural network coefficient representation. Graphs of recognition accuracy and the amount of FPGA hardware resources depending on the bit depth of neural network coefficient representation have been constructed. The coefficients obtained as a result of neural network training have been analyzed using decomposition into bit planes. It has been shown that 5 bits are sufficient to represent neural network coefficients, since they contain the main information learned by the network, ensuring economical use of FPGA resources and high recognition accuracy (92.4 %).https://doklady.bsuir.by/jour/article/view/4117neural networkhandwritten digit recognitionfully connected layermnistfpgabit planes
spellingShingle E. A. Krivalсevich
M. I. Vashkevich
Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
Doklady Belorusskogo gosudarstvennogo universiteta informatiki i radioèlektroniki
neural network
handwritten digit recognition
fully connected layer
mnist
fpga
bit planes
title Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
title_full Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
title_fullStr Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
title_full_unstemmed Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
title_short Investigation of Hardware Implementation of a Feedforward Neural Network for Handwritten Digit Recognition Based on FPGA
title_sort investigation of hardware implementation of a feedforward neural network for handwritten digit recognition based on fpga
topic neural network
handwritten digit recognition
fully connected layer
mnist
fpga
bit planes
url https://doklady.bsuir.by/jour/article/view/4117
work_keys_str_mv AT eakrivalsevich investigationofhardwareimplementationofafeedforwardneuralnetworkforhandwrittendigitrecognitionbasedonfpga
AT mivashkevich investigationofhardwareimplementationofafeedforwardneuralnetworkforhandwrittendigitrecognitionbasedonfpga