UTILIZATION OFGRAPHDATASTRUCTUREANDGRAPHNEURAL NETWORKSFORLIGHTWEIGHTANDEFFICIENTHUMANACTIVITY RECOGNITIONINE-HEALTHAPPLICATIONS

Date

2024-9-6

Author

Mursal, Yağmur

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

42
views

0
downloads

Cite This

Human activity recognition (HAR) plays a crucial role in applications like healthcare, and smart environments, aiming to improve health outcomes, and optimize daily living. Traditional HAR systems often rely on complex models unsuitable for edge computing due to high computational requirements. This thesis addresses the need for lightweight models that maintain high accuracy while being comptationally light. To achieve this, it proposes the use of graph neural networks (GNNs), a mechanism that has been extensively studied but is relatively underexplored for Human Activity Recognition (HAR) using various sensors, such as accelerometers, gyroscopes, and magnetometers. The proposed mechanism uses bidirectional Long Short-Term Memory (LSTM) layers for temporal feature extraction and GNNs for activity classification. The model’s performance is validated on a well known dataset through extensive experiments, demonstrating a detection accuracy of 93.19%, with significantly lower computational requirements compared to existing models. Sensor importance analysis highlights the critical role of gyroscope and accelerometer sensors in capturing detailed motion data.

Subject Keywords

e-health, graph neural network, lightweight, human activity recognition

URI

https://hdl.handle.net/11511/112992

Collections

Northern Cyprus Campus, Thesis