Citation

Prity, Fariya Sultana and Fahad, Nafiz and Ahmed, Rasel and Morol, Md. Kishor and Goh, Kah Ong Michael and Sadib, Ridwan Jamal and Hossen, Md. Jakir and Nawshad, Nadim and Hamid, Md Abdul and Sen, Anik and Rabbi, Riadul Islam and Rahman, Kazi Ashikur (2025) Advancements in human movement monitoring: A comprehensive approach leveraging deep learning and machine learning techniques. In: 4th International Conference on Computer, Information Technology and Intelligent Computing, CITIC 2024, 23 July 2024 - 25 July 2024, Virtual, Online.

Text conference.pdf - Published Version Restricted to Repository staff only Download (203kB)

Abstract

Using a novel dataset combined with a latest state-of-the-art deep learning and machine learning methods (including Residual Neural Networking (ResNet), k-nearest Neighbors (KNN), Decision Trees (DT), Logistic regression (LR), and Random Forests (RF)) this research introduced a cutting-edge approach in human motion tracking. The study primarily concerned with accurately identifying human movement in the context of religious postures, which has never been developed in scientific literature before. The approach involves the establishment of a novel dataset with 2000 images, considered seven diverse human religious prayer activities using CCTV cameras. This dataset was carefully collected to ensure diverse angles and lighting conditions, allowing for complete visibility and facilitating movement detection and classification. The dataset was pre-processed to convert text information into numbers and to fix missing data points for most suitable usage. We used several machines learning classifiers such as ResNet50, KNN, DT, LR, and RF to analyze and predict human movement accurately. It highlighted the best performance of various classifiers on this unique dataset and chose the best model for detecting human activity movements. The superior performance of the classifier ResNet50, achieving an accuracy of 99.6%, which is higher than previous studies in the literature, is the highlight of the research. The findings of this research have far-reaching consequences for the uses from sports science to health care and robotics where accurate human motion tracking is vital. In the future, we will work to augment the dataset and try a wider variety of deep-learning and traditional machine learning architectures. These models are good now, we also want to keep improving their accuracy & build an AI framework for a wide variety of use cases. These findings provide a strong foundation upon which future human movement analysis research can build, illustrating the promise of using advanced machine learning methods on domain-specific data.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Deep learning, artificial neural networks, machine learning
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Nor Afiqah Mohd Adnan
Date Deposited:	04 Dec 2025 08:41
Last Modified:	04 Dec 2025 08:41
URII:	http://shdl.mmu.edu.my/id/eprint/14963

Downloads

Edit (login required)