Physical Activity Recognition Based on a Parallel Approach for an Ensemble of Machine Learning and Deep Learning...

Area of Science:

• Biomedical Engineering
• Computer Science
• Wearable Technology

Background:

• Human Activity Recognition (HAR) using wearable sensors is crucial for remote health monitoring and emergency notification.
• The Internet of Things (IoT) enables seamless integration of wearable devices for enhanced healthcare standards.

Purpose of the Study:

• To investigate a human activity recognition method with improved accuracy and speed for healthcare applications.
• To develop a robust HAR system applicable in real-world healthcare scenarios.

Main Methods:

• A hybrid approach combining feature engineering and feature learning for data representation.
• Classification of wearable sensor acceleration time series data from human movement.
• Leave-one-subject-out cross-validation using data from 44 subjects wearing a waist-worn accelerometer.

Main Results:

• Achieved an average human activity recognition rate of 90%.
• Demonstrated significantly better performance compared to individual classification methods.
• The method supports functional and computational parallelization, reducing execution time.

Conclusions:

• The proposed HAR method is effective and efficient for healthcare applications.
• High accuracy and speed make it suitable for remote health monitoring and emergency notification.
• The approach offers a promising solution for advancing healthcare through wearable technology.