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...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-09-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/22/19/7324 |
| _version_ | 1797476963053469696 |
|---|---|
| author | Nuno Bento Joana Rebelo Marília Barandas André V. Carreiro Andrea Campagner Federico Cabitza Hugo Gamboa |
| author_facet | Nuno Bento Joana Rebelo Marília Barandas André V. Carreiro Andrea Campagner Federico Cabitza Hugo Gamboa |
| author_sort | Nuno Bento |
| collection | DOAJ |
| 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. |
| first_indexed | 2024-03-09T21:11:15Z |
| format | Article |
| id | doaj.art-8feb243e58774b089adfe924c0796b42 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T21:11:15Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-8feb243e58774b089adfe924c0796b422023-11-23T21:47:22ZengMDPI AGSensors1424-82202022-09-012219732410.3390/s22197324Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity RecognitionNuno Bento0Joana Rebelo1Marília Barandas2André V. Carreiro3Andrea Campagner4Federico Cabitza5Hugo Gamboa6Associação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, PortugalAssociação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, PortugalAssociação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, PortugalAssociação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, PortugalDipartimento di Informatica, Sistemistica e Comunicazione, Università degli Studi di Milano-Bicocca, 20126 Milan, ItalyDipartimento di Informatica, Sistemistica e Comunicazione, Università degli Studi di Milano-Bicocca, 20126 Milan, ItalyAssociação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, PortugalHuman 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.https://www.mdpi.com/1424-8220/22/19/7324human activity recognitiondeep learningdomain generalizationaccelerometer |
| spellingShingle | Nuno Bento Joana Rebelo Marília Barandas André V. Carreiro Andrea Campagner Federico Cabitza Hugo Gamboa Comparing Handcrafted Features and Deep Neural Representations for Domain Generalization in Human Activity Recognition Sensors human activity recognition deep learning domain generalization accelerometer |
| 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 | human activity recognition deep learning domain generalization accelerometer |
| url | https://www.mdpi.com/1424-8220/22/19/7324 |
| work_keys_str_mv | AT nunobento comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT joanarebelo comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT mariliabarandas comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT andrevcarreiro comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT andreacampagner comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT federicocabitza comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition AT hugogamboa comparinghandcraftedfeaturesanddeepneuralrepresentationsfordomaingeneralizationinhumanactivityrecognition |