Sitting Posture Monitoring System Based on a Low-Cost Load Cell Using Machine Learning

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Machine Learning Applications

Background:

• Existing sitting posture monitoring systems (SPMSs) often require numerous sensors on both seat and backrest plates.
• This limitation increases complexity and potentially cost for real-time posture assessment solutions.

Purpose of the Study:

• To develop a simplified SPMS with reduced sensor count for accurate real-time sitting posture classification.
• To evaluate the efficacy of machine learning algorithms in identifying sitting postures using load cell data.

Main Methods:

• Developed an SPMS utilizing four load cells exclusively mounted on the seat plate.
• Measured body weight ratios to classify six distinct sitting postures, including those with backrest loading.
• Applied and compared various machine learning algorithms for posture classification.

Main Results:

• A support vector machine with a radial basis function kernel achieved the highest classification rates (average 97.20%, maximum 97.94%) across nine subjects.
• This classifier demonstrated statistically significant improvements over other methods.
• The system successfully classified six sitting postures with a reduced sensor configuration.

Conclusions:

• The proposed SPMS effectively classifies multiple sitting postures using a minimal sensor setup on the seat plate.
• This approach offers a viable, potentially more accessible, alternative to complex, multi-sensor SPMS.
• Reduced sensor count in SPMS is feasible for accurate real-time posture monitoring.