Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors

With the increasing availability of wearable devices for data collection, studies in human activity recognition have gained significant popularity. These studies report high accuracies on k-fold cross validation, which is not reflective of their generalization performance but is a result of the inap...

Full description

Saved in:
Bibliographic Details
Main Authors: Saeed Ur Rehman, Anwar Ali, Adil Mehmood Khan, Cynthia Okpala
Format: Article
Language:English
Published: MDPI AG 2024-12-01
Series:Algorithms
Subjects:
Online Access:https://www.mdpi.com/1999-4893/17/12/556
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846106308546134016
author Saeed Ur Rehman
Anwar Ali
Adil Mehmood Khan
Cynthia Okpala
author_facet Saeed Ur Rehman
Anwar Ali
Adil Mehmood Khan
Cynthia Okpala
author_sort Saeed Ur Rehman
collection DOAJ
description With the increasing availability of wearable devices for data collection, studies in human activity recognition have gained significant popularity. These studies report high accuracies on k-fold cross validation, which is not reflective of their generalization performance but is a result of the inappropriate split of testing and training datasets, causing these models to evaluate the same subjects that they were trained on, making them subject-dependent. This study comparatively discusses this validation approach with a universal approach, Leave-One-Subject-Out (LOSO) cross-validation which is not subject-dependent and ensures that an entirely new subject is used for evaluation in each fold, validated on four different machine learning models trained on windowed data and select hand-crafted features. The random forest model, with the highest accuracy of 76% when evaluated on LOSO, achieved an accuracy of 89% on k-fold cross-validation, demonstrating data leakage. Additionally, this experiment underscores the significance of hand-crafted features by contrasting their accuracy with that of raw sensor models. The feature models demonstrate a remarkable 30% higher accuracy, underscoring the importance of feature engineering in enhancing the robustness and precision of HAR systems.
format Article
id doaj-art-e2308b02075d406ca3548f159c93f454
institution Kabale University
issn 1999-4893
language English
publishDate 2024-12-01
publisher MDPI AG
record_format Article
series Algorithms
spelling doaj-art-e2308b02075d406ca3548f159c93f4542024-12-27T14:05:13ZengMDPI AGAlgorithms1999-48932024-12-01171255610.3390/a17120556Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable SensorsSaeed Ur Rehman0Anwar Ali1Adil Mehmood Khan2Cynthia Okpala3Faculty of Science and Engineering, University of Hull, Cottingham Rd., Hull HU6 7RX, UKDepartment of Electronic and Electrical Engineering, Swansea University Bay Campus, Swansea SA1 8EN, UKFaculty of Science and Engineering, University of Hull, Cottingham Rd., Hull HU6 7RX, UKFaculty of Science and Engineering, University of Hull, Cottingham Rd., Hull HU6 7RX, UKWith the increasing availability of wearable devices for data collection, studies in human activity recognition have gained significant popularity. These studies report high accuracies on k-fold cross validation, which is not reflective of their generalization performance but is a result of the inappropriate split of testing and training datasets, causing these models to evaluate the same subjects that they were trained on, making them subject-dependent. This study comparatively discusses this validation approach with a universal approach, Leave-One-Subject-Out (LOSO) cross-validation which is not subject-dependent and ensures that an entirely new subject is used for evaluation in each fold, validated on four different machine learning models trained on windowed data and select hand-crafted features. The random forest model, with the highest accuracy of 76% when evaluated on LOSO, achieved an accuracy of 89% on k-fold cross-validation, demonstrating data leakage. Additionally, this experiment underscores the significance of hand-crafted features by contrasting their accuracy with that of raw sensor models. The feature models demonstrate a remarkable 30% higher accuracy, underscoring the importance of feature engineering in enhancing the robustness and precision of HAR systems.https://www.mdpi.com/1999-4893/17/12/556machine learningLOSOhuman activity recognition
spellingShingle Saeed Ur Rehman
Anwar Ali
Adil Mehmood Khan
Cynthia Okpala
Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
Algorithms
machine learning
LOSO
human activity recognition
title Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
title_full Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
title_fullStr Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
title_full_unstemmed Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
title_short Human Activity Recognition: A Comparative Study of Validation Methods and Impact of Feature Extraction in Wearable Sensors
title_sort human activity recognition a comparative study of validation methods and impact of feature extraction in wearable sensors
topic machine learning
LOSO
human activity recognition
url https://www.mdpi.com/1999-4893/17/12/556
work_keys_str_mv AT saeedurrehman humanactivityrecognitionacomparativestudyofvalidationmethodsandimpactoffeatureextractioninwearablesensors
AT anwarali humanactivityrecognitionacomparativestudyofvalidationmethodsandimpactoffeatureextractioninwearablesensors
AT adilmehmoodkhan humanactivityrecognitionacomparativestudyofvalidationmethodsandimpactoffeatureextractioninwearablesensors
AT cynthiaokpala humanactivityrecognitionacomparativestudyofvalidationmethodsandimpactoffeatureextractioninwearablesensors