"A novel adaptive gesture recognition framework for bionic hands using Stacked Autoencoder (SAE), Adaptive Bayesian...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Rehabilitation Technology

Background:

• Prosthetic limbs often lack intuitive control, hindering natural movement for amputees.
• Existing systems may not fully capture the nuances of residual limb muscle signals.
• There is a need for advanced prosthetic control that integrates seamlessly with user intent.

Purpose of the Study:

• To develop and evaluate an intelligent bionic hand system for intuitive prosthetic control.
• To explore the combination of muscle magnetic sensing (MMG) and surface electromyography (sEMG) for enhanced signal acquisition.
• To create a self-learning prosthetic hand that adapts to individual user muscle patterns.

Main Methods:

• Integration of muscle magnetic sensing (MMG) and surface electromyography (sEMG) for comprehensive muscle activity detection.
• Utilization of advanced neural networks for a self-learning AI system that continuously refines muscle pattern recognition.
• Real-world testing with 15 participants, including ADAMS-MATLAB co-simulation and 3D-printed prototype evaluation.

Main Results:

• The bionic hand demonstrated high precision in capturing subtle muscle movements.
• The system achieved up to 99.9% accuracy in subject-specific cross-validation with a processing delay as low as 12 ms.
• Preliminary hardware experiments and simulations confirmed the system's strong feasibility for real-world application.

Conclusions:

• The developed intelligent bionic hand offers a personalized and adaptive solution for prosthetic control.
• This technology bridges the gap between human biological signals and machine function, promising more natural limb movement.
• Further large-scale and long-term studies are warranted, but this work represents a significant step towards next-generation prosthetics that restore function and confidence.