A decision-based velocity ramp for minimizing the effect of misclassifications during real-time pattern recognition...

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Human-Computer Interaction

Background:

• Real-time pattern recognition control for prosthetics is often hindered by classification errors.
• Unintended movements resulting from misclassifications can negatively impact prosthesis control and user performance.

Purpose of the Study:

• To investigate the efficacy of a decision-based velocity ramp in mitigating the effects of misclassifications in pattern recognition control.
• To enhance prosthesis positioning and minimize unintended movements during real-time control.

Main Methods:

• Implemented a decision-based velocity ramp that reduced movement speed following a classifier decision change.
• Evaluated the system with non-amputee and amputee subjects performing target achievement tests in virtual and physical environments.
• Compared performance with and without the velocity ramp, and against a majority vote strategy.

Main Results:

• Subjects demonstrated significantly higher completion rates and more direct paths to targets in virtual environments with the velocity ramp (p < 0.05).
• Participants using a physical prosthesis stacked significantly more cubes (76-89% increase) in three minutes with the velocity ramp (p < 0.05).
• The velocity ramp outperformed majority vote and was preferred by 83% of subjects.

Conclusions:

• A decision-based velocity ramp effectively improves real-time pattern recognition control for prosthetic devices.
• This post-processing technique offers a versatile solution for enhancing user performance across various classifiers and applications.
• The velocity ramp shows significant potential for improving prosthetic functionality and user experience.