Loading Recognition for Lumbar Exoskeleton Based on Multi-Channel Surface Electromyography From Low Back Muscles.

Area of Science:

• Biomechanics
• Robotics
• Human-Computer Interaction

Background:

• Lumbar exoskeletons assist with heavy lifting, but optimizing their support to minimize muscle energy consumption remains a challenge.
• Accurate recognition of low back muscle loading is crucial for adaptive exoskeleton assistance but is limited by current measurement tools and classification methods.

Purpose of the Study:

• To precisely identify muscle loading in the low back during lifting tasks.
• To develop and evaluate a participant-specific load classification method using multi-channel surface electromyography (sEMG).

Main Methods:

• Ten healthy participants performed stoop lifting with varying weights while sEMG data was collected from the low back using a 3x7 electrode array.
• Time-domain sEMG features were extracted from 19 lifting phase segments, and participant-specific classifiers were trained using four algorithms.
• Classification performance was assessed via 5-fold cross-validation.

Main Results:

• An artificial neural network classifier achieved up to 96% accuracy in identifying object weight during the lifting task.
• Classification accuracy improved as the lifting phase progressed, peaking towards the end of the movement.
• The study demonstrated successful, accurate recognition of low back muscle loading.

Conclusions:

• Accurate load recognition using sEMG and participant-specific classifiers is feasible.
• This method holds significant potential for reducing muscle energy consumption in lumbar exoskeleton users.
• The findings pave the way for more effective and energy-efficient assistive robotic devices.