A Spatio-Temporal Graph Convolutional Network Model for Internet of Medical Things (IoMT).

Area of Science:

• Computer Science
• Biomedical Engineering
• Artificial Intelligence

Background:

• Intelligent healthcare services in the Internet of Medical Things (IoMT) require efficient human action recognition (HAR).
• Classical HAR techniques face challenges with computational complexity and memory efficiency, limiting their applicability in IoMT.
• There is a need for novel HAR methods optimized for resource-constrained IoMT environments.

Purpose of the Study:

• To introduce a novel HAR technique, the spatio-temporal graph convolutional network (STGCN), for healthcare services within the IoMT.
• To address the limitations of traditional HAR methods in terms of computational complexity and memory usage.
• To develop a skeleton-based HAR model suitable for human-machine interfaces in intelligent healthcare.

Main Methods:

• The proposed method utilizes a spatio-temporal graph convolutional network (STGCN) for skeleton-based HAR.
• Spatial and temporal features are extracted independently to minimize information loss.
• The model focuses on using only joint data and a reduced number of parameters.

Main Results:

• The STGCN model achieved 92.2% accuracy on a skeleton dataset.
• The proposed method demonstrates superior performance compared to multi-channel methods that use both joint and bone data.
• STGCN offers a favorable balance between accuracy, memory consumption, and processing time.

Conclusions:

• The STGCN model is a viable and efficient solution for HAR in IoMT healthcare services.
• Its ability to balance accuracy and resource efficiency makes it suitable for detecting medical conditions.
• The independent extraction of spatio-temporal features ensures the capture of essential HAR information.