Cargando…

Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition

Human Activity Recognition (HAR) has been studied extensively, yet current approaches are not capable of generalizing across different domains (i.e., subjects, devices, or datasets) with acceptable performance. This lack of generalization hinders the applicability of these models in real-world envir...

Descripción completa

Detalles Bibliográficos
Autores principales: Bento, Nuno, Rebelo, Joana, Barandas, Marília, Carreiro, André V., Campagner, Andrea, Cabitza, Federico, Gamboa, Hugo
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9572241/
https://www.ncbi.nlm.nih.gov/pubmed/36236427
http://dx.doi.org/10.3390/s22197324
_version_ 1784810564091904000
author Bento, Nuno
Rebelo, Joana
Barandas, Marília
Carreiro, André V.
Campagner, Andrea
Cabitza, Federico
Gamboa, Hugo
author_facet Bento, Nuno
Rebelo, Joana
Barandas, Marília
Carreiro, André V.
Campagner, Andrea
Cabitza, Federico
Gamboa, Hugo
author_sort Bento, Nuno
collection PubMed
description Human Activity Recognition (HAR) has been studied extensively, yet current approaches are not capable of generalizing across different domains (i.e., subjects, devices, or datasets) with acceptable performance. This lack of generalization hinders the applicability of these models in real-world environments. As deep neural networks are becoming increasingly popular in recent work, there is a need for an explicit comparison between handcrafted and deep representations in Out-of-Distribution (OOD) settings. This paper compares both approaches in multiple domains using homogenized public datasets. First, we compare several metrics to validate three different OOD settings. In our main experiments, we then verify that even though deep learning initially outperforms models with handcrafted features, the situation is reversed as the distance from the training distribution increases. These findings support the hypothesis that handcrafted features may generalize better across specific domains.
format Online
Article
Text
id pubmed-9572241
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-95722412022-10-17 Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition Bento, Nuno Rebelo, Joana Barandas, Marília Carreiro, André V. Campagner, Andrea Cabitza, Federico Gamboa, Hugo Sensors (Basel) Article Human Activity Recognition (HAR) has been studied extensively, yet current approaches are not capable of generalizing across different domains (i.e., subjects, devices, or datasets) with acceptable performance. This lack of generalization hinders the applicability of these models in real-world environments. As deep neural networks are becoming increasingly popular in recent work, there is a need for an explicit comparison between handcrafted and deep representations in Out-of-Distribution (OOD) settings. This paper compares both approaches in multiple domains using homogenized public datasets. First, we compare several metrics to validate three different OOD settings. In our main experiments, we then verify that even though deep learning initially outperforms models with handcrafted features, the situation is reversed as the distance from the training distribution increases. These findings support the hypothesis that handcrafted features may generalize better across specific domains. MDPI 2022-09-27 /pmc/articles/PMC9572241/ /pubmed/36236427 http://dx.doi.org/10.3390/s22197324 Text en © 2022 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Bento, Nuno
Rebelo, Joana
Barandas, Marília
Carreiro, André V.
Campagner, Andrea
Cabitza, Federico
Gamboa, Hugo
Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title_full Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title_fullStr Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title_full_unstemmed Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title_short Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition
title_sort comparing handcrafted features and deep neural representations for domain generalization in human activity recognition
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9572241/
https://www.ncbi.nlm.nih.gov/pubmed/36236427
http://dx.doi.org/10.3390/s22197324
work_keys_str_mv AT bentonuno comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT rebelojoana comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT barandasmarilia comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT carreiroandrev comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT campagnerandrea comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT cabitzafederico comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition
AT gamboahugo comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition