Smartphone-Based Human Sitting Behaviors Recognition Using Inertial Sensor.

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Wearable Technology

Background:

• Prolonged sitting is linked to adverse health outcomes like shoulder pain and headaches.
• Maintaining dynamic postures is crucial for mitigating health risks associated with sedentary office work.
• Current methods for monitoring human sitting behaviors are often limited or invasive.

Purpose of the Study:

• To develop and evaluate a smartphone-based system for detecting unhealthy human sitting behaviors (HSBs) in office workers.
• To investigate the efficacy of inertial sensors for real-time posture monitoring.
• To identify optimal feature selection and machine learning techniques for accurate HSB classification.

Main Methods:

• Utilized inertial sensors (accelerometer, gyroscope, magnetometer) integrated into smartphones.
• Collected data from six volunteers performing five distinct sitting activities.
• Employed Correlation-based Feature Selection (CFS) and Particle Swarm Optimization (PSO) for feature vector optimization.
• Trained and evaluated supervised machine learning classifiers: Naive Bayes, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN).

Main Results:

• The Support Vector Machine (SVM) classifier demonstrated superior performance, achieving 99.90% overall accuracy.
• Optimized features derived from CFS and PSO significantly improved classification accuracy.
• The system effectively distinguished between various human sitting behaviors using smartphone sensor data.

Conclusions:

• Smartphone-integrated inertial sensors offer a viable and accurate solution for monitoring human sitting behaviors.
• Machine learning, particularly SVM, combined with optimized features, is highly effective for recognizing HSBs.
• This technology has the potential to promote healthier sitting habits and reduce associated health risks for office workers.