Rejection of Systemic and Operator Errors in a Real-Time Myoelectric Control Task

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Human-Computer Interaction

Background:

• Myoelectric pattern recognition improves prosthetic control.
• Rejecting low-confidence decisions enhances usability.
• The impact of rejection on error mitigation versus user adaptation is unclear.

Purpose of the Study:

• To differentiate the effects of error mitigation and user adaptation in myoelectric control.
• To quantify the influence of confidence thresholds on systemic and operator errors.
• To assess the trade-off between error reduction and filtering of correct decisions.

Main Methods:

• 24 subjects performed real-time pattern recognition tasks.
• Rejection was applied at seven confidence thresholds and in a no-rejection condition.
• Errors were categorized as systemic (classifier) or operator (user).

Main Results:

• High rejection thresholds halved overall error rates.
• Both systemic and operator errors were significantly reduced.
• User-produced errors remained constant, indicating rejection primarily affects classifier errors.
• Correct decisions were increasingly filtered at higher rejection thresholds.

Conclusions:

• Myoelectric control rejection effectively mitigates classifier errors.
• Rejection does not significantly alter user error patterns.
• Excessive rejection may hinder usability by filtering correct commands.
• Experience level did not influence rejection effectiveness.