Context-driven, prescription-based personal activity classification: methodology, architecture, and end-to-end...

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human-Computer Interaction

Background:

• Large-scale monitoring of range of motion (ROM) activities is crucial for healthcare and fitness professionals to ensure exercise quality and compliance.
• Existing methods face challenges in scalability, limiting their widespread application.
• There is a need for robust systems that can accurately classify diverse ROM activities in real-world settings.

Purpose of the Study:

• To present a novel end-to-end system for scalable, context-driven classification of range of motion activities.
• To refine the prescription-based methodology by incorporating context and scenarios for personalized monitoring.
• To demonstrate a flexible architecture enabling efficient and accurate activity classification.

Main Methods:

• Developed a prescription-based, context-driven activity classification methodology.
• Introduced refined definitions of context and the concept of scenarios for personalized monitoring.
• Implemented a flexible architecture with interface models, a classification committee for context classification, and context-specific models for activity classification.
• Created an end-to-end system with an Android application and mobile device integration.

Main Results:

• The proposed system demonstrated improved classification accuracy and rate for ROM activities.
• Field evaluations confirmed the system's effectiveness in real-world scenarios.
• The approach contributed to extending sensor operating life through optimized monitoring.

Conclusions:

• The novel end-to-end system effectively addresses scalability challenges in ROM activity monitoring.
• The context-driven approach enhances personalization and accuracy in exercise quality and compliance assessment.
• The system offers a viable solution for large-scale deployment in healthcare and fitness applications.