Enhancing surface electromyographic signal recognition accuracy for trans-radial amputees using broad learning...

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) is vital for human-computer interaction, enabling control of myoelectric prostheses by analyzing residual muscle activity in trans-radial amputees.
• Existing gesture classification methods face accuracy challenges as the number of hand movement types increases.

Purpose of the Study:

• To develop a novel and accurate gesture recognition method for trans-radial amputees.
• To enhance the control of myoelectric prostheses through improved hand movement intention prediction.

Main Methods:

• A new approach integrating Image Feature Flattening (IFF) with a Broad Learning System (BLS) was proposed.
• The IFF method transforms sEMG data features into a 3D image, converts it to grayscale, and flattens it into a 1D vector for BLS input.
• The method was validated using the Ninapro DB3 dataset, comprising 49 hand movement types from amputees.

Main Results:

• The proposed method achieved a high gesture recognition accuracy of up to 98.1%.
• Significant reduction in classification time was observed compared to previous methods.
• The approach demonstrated strong potential for practical application in real-time gesture recognition.

Conclusions:

• The integration of IFF and BLS offers a highly accurate and efficient solution for gesture recognition in trans-radial amputees.
• This advancement holds significant promise for improving the functionality and user experience of myoelectric prostheses.
• The developed method effectively addresses the challenge of decreasing accuracy with an increasing number of gesture classes.