Application of Machine Learning Approaches for Classifying Sitting Posture Based on Force and Acceleration Sensors

Area of Science:

• Ergonomics and Occupational Health
• Biomedical Engineering
• Machine Learning Applications

Background:

• Prolonged static or nonergonomic sitting contributes to occupational musculoskeletal disorders, especially chronic low back pain (LBP).
• Accurate assessment of sitting positions is crucial for understanding and mitigating these health issues.

Purpose of the Study:

• To develop an instrumented chair equipped with force and acceleration sensors.
• To evaluate the accuracy of machine learning (ML) methods in automatically identifying user sitting positions.

Main Methods:

• Utilized an instrumented chair with 16 force sensors and a backrest angle sensor.
• Collected 1148 sitting samples from 41 subjects across seven distinct sitting positions.
• Applied and compared five ML algorithms: Support Vector Machines, Multinomial Regression, Boosting, Neural Networks, and Random Forest, using Leave-One-Out cross-validation.

Main Results:

• The Random Forest algorithm achieved the highest classification accuracy at 90.9% for unfamiliar subjects.
• Sitting position classification accuracy ranged from 81% to 98% across the seven tested positions.
• Demonstrated the feasibility of accurately classifying sitting postures using the instrumented chair and ML.

Conclusions:

• An instrumented office chair combined with ML analysis can accurately classify different sitting positions.
• This technology offers potential insights into the links between sitting behavior, posture, and musculoskeletal disorder development.
• Novel approaches like this can advance the assessment of chair usage and promote healthier sitting habits.