An Efficient Service-based System for Hierarchical Human Activity Sensing.

Area of Science:

• Mobile Health
• Edge Computing
• Human Activity Recognition

Background:

• Measuring Activities of Daily Living (ADL) is crucial for assessing elderly independence and rehabilitation progress.
• Existing methods often incur high data transfer costs and CPU usage.
• There is a need for efficient, on-device processing of continuous mobility data.

Purpose of the Study:

• To propose an end-to-end system for processing client edge-device mobility data.
• To optimize on-device processing pipelines for reduced data transfer and CPU load.
• To develop a classification algorithm for recognizing Activities of Daily Living (ADL).

Main Methods:

• Implementation of a SEnsing as Service (SEAS) model for continuous mobility data processing.
• Optimization of on-device processing pipelines on edge devices.
• Integration of a novel classification algorithm for ADL recognition.

Main Results:

• The proposed system effectively processes continuous mobility data at the client edge.
• Significant reduction in data transfer costs and CPU usage was achieved.
• The system demonstrated considerable accuracy and flexibility in recognizing Activities of Daily Living (ADL).

Conclusions:

• The developed SEAS-based system offers an efficient solution for on-device mobility data analysis.
• This approach enhances the feasibility of real-time ADL monitoring for healthcare applications.
• The system shows promise for supporting independent living and post-event rehabilitation.