Research on Intelligent Wheelchair Attitude-Based Adjustment Method Based on Action Intention Recognition.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Intelligent wheelchair research often overlooks posture adjustment, with existing methods lacking collaborative control.
• Current posture adjustment techniques in wheelchairs have limitations in human-machine collaboration.

Purpose of the Study:

• To propose an intelligent wheelchair posture-adjustment method based on recognizing user action intentions.
• To improve human-machine collaboration in intelligent wheelchairs through intuitive posture control.

Main Methods:

• Utilized a multi-part adjustable electric wheelchair equipped with force sensors to capture body pressure data.
• Developed a system to convert pressure data into distribution maps and employed the VIT deep learning model for feature extraction and intention classification.
• Controlled electric actuators to adjust wheelchair posture based on recognized action intentions.

Main Results:

• Successfully collected passenger body pressure data.
• Achieved over 95% accuracy in identifying common user intentions (lying down, sitting up, standing up).
• Demonstrated the wheelchair's ability to adjust posture according to recognized intentions.

Conclusions:

• The proposed method effectively collects body pressure data and accurately identifies user intentions for posture adjustment.
• This approach enhances wheelchair usability by enabling natural posture changes without external equipment, improving user independence and collaboration.