Simultaneous Recognition of Locomotion Mode, Phase, and Phase Progression Using Deep Learning Models.

Area of Science:

• Robotics
• Biomechanics
• Machine Learning

Background:

• Gait-assist wearable robots face limited real-world application due to challenges in user intention recognition.
• Effective algorithms are vital for robots to synchronize with users during diverse locomotion activities like walking and stair negotiation.

Purpose of the Study:

• To develop a deep learning framework for simultaneous prediction of locomotion modes, phases, and phase progression.
• To enhance personalized assistance in gait-assist wearable robots by accurately interpreting user intent.

Main Methods:

• Utilized a deep learning framework for automatic feature extraction from inertial measurement unit (IMU) data.
• Employed sensors placed on the sternum and limbs to capture comprehensive movement data.
• Evaluated model generalizability using a leave-one-subject-out cross-validation approach.

Main Results:

• The deep learning model effectively classified locomotion phases.
• The framework accurately estimated the percentage of phase progression.
• The leave-one-subject-out evaluation demonstrated robust generalizability across different users.

Conclusions:

• The proposed deep learning framework offers a significant advancement in user intention recognition for gait-assist robots.
• Simultaneous prediction of locomotion mode, phase, and progression enables more harmonious and personalized robotic assistance.
• The study highlights the potential of IMU-based deep learning for real-world deployment of assistive robotic technologies.