Intelligent Sitting Posture Classifier for Wheelchair Users

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Rehabilitation Technology

Background:

• Growing interest in seated postural monitoring to prevent long-term ulcers and musculoskeletal issues.
• Current subjective questionnaires lack continuous, quantitative postural data for wheelchair users.
• Need for objective monitoring to track postural status and disease-related anomalies.

Purpose of the Study:

• To propose an intelligent classifier for classifying sitting postures of wheelchair users.
• To develop a system for continuous and quantitative postural monitoring.
• To support wheelchair users and healthcare professionals in automatic posture management.

Main Methods:

• Development of an intelligent classifier based on a multilayer neural network.
• Posture database generated using data from a novel force resistive sensor monitoring device.
• Training and hyperparameter selection using a stratified K-Fold in weight groups strategy for enhanced generalization.

Main Results:

• The proposed neural network model achieved higher success rates compared to other models.
• The system demonstrates improved generalization capabilities, performing well on subjects with diverse physical complexions.
• Successful classification of sitting postures for wheelchair users, regardless of physical variations.

Conclusions:

• The intelligent classifier provides an effective solution for automatic posture monitoring in wheelchair users.
• The system enhances the capacity for generalization, accommodating a wide range of physical complexions.
• This technology can significantly aid in preventing long-term health complications associated with poor posture.