Enhancing Locomotion-Mode Recognition and Transition Prediction With (Bio)Mechanical Sensor Fusion for Intelligent...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Prosthetics

Background:

• Intelligent prosthetic knees require continuous locomotion recognition and accurate transition prediction.
• Current methods face challenges in real-time accuracy and efficiency.

Purpose of the Study:

• To develop and validate an innovative framework for locomotion recognition and transition prediction in prosthetic knees.
• To fuse mechanical (IMU) and biomechanical (FMG) signals for enhanced performance.

Main Methods:

• An FMG-IMU dual-modal sensing system was implemented on a prosthetic knee.
• A feature-driven CNN-BiLSTM model was developed for classification.
• The RelifF-MI algorithm was used for feature optimization.

Main Results:

• The FMG-IMU fusion framework achieved an average recognition accuracy of 98.51%.
• Transition prediction was achieved with an average time of 274 ms (21.82% of gait cycle).
• Significant improvement in classification performance was observed across five locomotion modes and eight transitions.

Conclusions:

• The proposed method demonstrates high accuracy and efficiency for prosthetic knee control.
• Fusion of FMG-IMU signals with a feature-driven classifier holds significant potential for advanced prosthetic applications.