Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module

The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operati...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Mateusz Chmurski, Gianfranco Mauro, Avik Santra, Mariusz Zubert, Gökberk Dagasan
Formato: article
Lenguaje:EN
Publicado: MDPI AG 2021
Materias:
Acceso en línea:https://doaj.org/article/f11070adb7264dbda7fdd7902e00ba6e
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:f11070adb7264dbda7fdd7902e00ba6e
record_format dspace
spelling oai:doaj.org-article:f11070adb7264dbda7fdd7902e00ba6e2021-11-11T19:15:03ZHighly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module10.3390/s212172981424-8220https://doaj.org/article/f11070adb7264dbda7fdd7902e00ba6e2021-11-01T00:00:00Zhttps://www.mdpi.com/1424-8220/21/21/7298https://doaj.org/toc/1424-8220The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operating environment. Further, the deployment of such systems is often not performed in resource-constrained contexts. Inspired by the MobileNetV1 deep learning network, this paper presents a novel hand gesture recognition system based on frequency-modulated continuous wave (FMCW) radar, exhibiting a higher recognition accuracy in comparison to the state-of-the-art systems. First of all, the paper introduces a method to simplify radar preprocessing while preserving the main information of the performed gestures. Then, a deep neural classifier with the novel Depthwise Expansion Module based on the depthwise separable convolutions is presented. The introduced classifier is optimized and deployed on the Coral Edge TPU board. The system defines and adopts eight different hand gestures performed by five users, offering a classification accuracy of 98.13% while operating in a low-power and resource-constrained environment.Mateusz ChmurskiGianfranco MauroAvik SantraMariusz ZubertGökberk DagasanMDPI AGarticleedge computingEdge TPUoptimizationquantizationFMCWradarChemical technologyTP1-1185ENSensors, Vol 21, Iss 7298, p 7298 (2021)
institution DOAJ
collection DOAJ
language EN
topic edge computing
Edge TPU
optimization
quantization
FMCW
radar
Chemical technology
TP1-1185
spellingShingle edge computing
Edge TPU
optimization
quantization
FMCW
radar
Chemical technology
TP1-1185
Mateusz Chmurski
Gianfranco Mauro
Avik Santra
Mariusz Zubert
Gökberk Dagasan
Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
description The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operating environment. Further, the deployment of such systems is often not performed in resource-constrained contexts. Inspired by the MobileNetV1 deep learning network, this paper presents a novel hand gesture recognition system based on frequency-modulated continuous wave (FMCW) radar, exhibiting a higher recognition accuracy in comparison to the state-of-the-art systems. First of all, the paper introduces a method to simplify radar preprocessing while preserving the main information of the performed gestures. Then, a deep neural classifier with the novel Depthwise Expansion Module based on the depthwise separable convolutions is presented. The introduced classifier is optimized and deployed on the Coral Edge TPU board. The system defines and adopts eight different hand gestures performed by five users, offering a classification accuracy of 98.13% while operating in a low-power and resource-constrained environment.
format article
author Mateusz Chmurski
Gianfranco Mauro
Avik Santra
Mariusz Zubert
Gökberk Dagasan
author_facet Mateusz Chmurski
Gianfranco Mauro
Avik Santra
Mariusz Zubert
Gökberk Dagasan
author_sort Mateusz Chmurski
title Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
title_short Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
title_full Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
title_fullStr Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
title_full_unstemmed Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module
title_sort highly-optimized radar-based gesture recognition system with depthwise expansion module
publisher MDPI AG
publishDate 2021
url https://doaj.org/article/f11070adb7264dbda7fdd7902e00ba6e
work_keys_str_mv AT mateuszchmurski highlyoptimizedradarbasedgesturerecognitionsystemwithdepthwiseexpansionmodule
AT gianfrancomauro highlyoptimizedradarbasedgesturerecognitionsystemwithdepthwiseexpansionmodule
AT aviksantra highlyoptimizedradarbasedgesturerecognitionsystemwithdepthwiseexpansionmodule
AT mariuszzubert highlyoptimizedradarbasedgesturerecognitionsystemwithdepthwiseexpansionmodule
AT gokberkdagasan highlyoptimizedradarbasedgesturerecognitionsystemwithdepthwiseexpansionmodule
_version_ 1718431620532273152