Position-independent human activity recognition for its using smartphone sensors : comparative analysis of machine learning and deep learning approaches | |
| Author | Bernardo, John Benedict L. |
| Call Number | AIT Diss. no.ICT-24-01 |
| Subject(s) | Human activity recognition Machine learning Deep learning (Machine learning) |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Information and Communication Technologies |
| Publisher | Asian Institute of Technology |
| Abstract | This study nvestigates Human Activity Recognition (HAR )usin gsmartphone-embedded sensors to address limitations associated with position-dependent datasets within Intel ligent Transportation Systems (ITS). A position-independent framework is proposed, utilizing data from accelerometers, gyroscopes, linear accelerometers, and gravity sen sors, with smartphone placements on the chest or in left/right leg pockets. The per formance of traditional machine learning algorithms—Decision Trees (DT), K-Nearest Neighbors (KNN), Random Forest (RF), Support Vector Classifier (SVC), and XG Boost—is comparedagainst advanced deep learning models, including Gated Recurrent Units (GRU), Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), and Transformers, under two sensor configurations. Results indicate that the Temporal Convolutional Network (TCN) model consistently achieves superior performance, with an accuracy of 97.70% in the 4-sensor non over lapping configuration. Deep learning models, particularly LSTM, GRU, and Trans former, exhibit strong capability in capturing temporal dependencies in HAR tasks, while traditional machine learning models, such as RF and XGBoost, demonstrate rea sonable accuracy but do not match the efficacy of deep learning approaches. Further, incorporating data from linear accelerometers and gravity sensors yields modest accu racy improvements over configurations limited to accelerometers and gyroscopes alone, enhancing the robustness of activity recognition. These findings contribute to the advancement of HAR in ITS by improving passenger activity recognition, which is essential for optimizing congestion management, safety protocols, and emergency response strategies in dynamic transportation environments. |
| Year | 2024 |
| Type | Dissertation |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information and Communication Technology (ICT) |
| Chairperson(s) | Attaphongse Taparugssanagorn |
| Examination Committee(s) | Chaklam Silpasuwanchai;Chantri Polprasert |
| Scholarship Donor(s) | University of Science and Technology of Southern Philippines (USTP) |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2024 |