A Random Forest-based ensemble method for activity recognition.

Area of Science:

• Computer Science
• Biomedical Engineering
• Machine Learning

Background:

• Human physical activity (PA) recognition is crucial for health monitoring.
• Existing methods often face challenges with accuracy and computational efficiency.
• Developing robust and fast PA recognition systems is an ongoing research area.

Purpose of the Study:

• To present a novel multi-sensor ensemble approach for human PA recognition.
• To enhance the accuracy, stability, and speed of activity classification.
• To leverage Random Forest (RF) for an integrated classifier system.

Main Methods:

• An ensemble learning algorithm integrating multiple independent Random Forest (RF) classifiers was designed.
• Different sensor feature sets were utilized for each RF classifier.
• The Physical Activity Monitoring for Aging People (PAMAP) database was used for training and testing.

Main Results:

• The proposed ensemble classifier achieved a high recognition accuracy of 93.44% for 19 PA types.
• The system demonstrated faster training times compared to other methods.
• The multi-sensor ensemble approach proved effective in improving classification performance.

Conclusions:

• The ensemble classifier system based on the RF algorithm offers high recognition accuracy and computational efficiency.
• This approach provides a stable and accurate solution for human activity recognition.
• The findings suggest the potential of ensemble learning for advanced PA monitoring systems.