Activity recognition with smartphone support.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human Activity Recognition

Background:

• Accurate human activity recognition is crucial for health monitoring and rehabilitation.
• Existing methods often rely on complex sensor setups or limited activity types.

Purpose of the Study:

• To develop and validate a custom mobility classifier for accurate human activity detection.
• To compare the custom classifier's performance against established machine learning algorithms.

Main Methods:

• Utilized a smartphone accelerometer paired with a chest sensor for data collection.
• Designed, implemented, and validated a custom mobility classifier.
• Conducted offline analysis comparing the custom classifier with C4.5, CART, SVM, Multi-Layer Perceptrons, and Naïve Bayes algorithms.
• Performed trials in Ireland (N=6) and the Netherlands (N=24), analyzing 1165 minutes of recorded activities.

Main Results:

• The custom mobility classifier achieved high accuracy, reaching up to 98% in recognizing activities.
• The system demonstrated capability in distinguishing between activities such as sitting, standing, lying, walking, running, and cycling.
• The custom classifier showed competitive or superior performance compared to traditional machine learning algorithms.

Conclusions:

• Smartphone accelerometers combined with chest sensors offer a viable and accurate method for human activity recognition.
• The developed custom mobility classifier provides a robust tool for detailed mobility analysis.
• This technology has potential applications in healthcare, sports science, and personal well-being monitoring.