Cross-day analysis of Multicode Surface Electromyography based Biometrics for Personal Identification

Area of Science:

• Biometrics
• Signal Processing
• Human-Computer Interaction

Background:

• Surface electromyography (sEMG) is a novel biometric trait for personal identification.
• sEMG offers dual-mode security: individual uniqueness (biometric-mode) and gesture-specific patterns (knowledge-mode).
• Previous studies on sEMG multicode frameworks were limited by single-day data and small participant groups.

Purpose of the Study:

• To evaluate the effectiveness of a multicode surface electromyography (sEMG) biometric framework for personal identification.
• To investigate the performance of sEMG-based identification in a cross-day scenario.
• To determine the optimal fusion scheme (score, rank, or decision) for multicode sEMG biometrics.

Main Methods:

• Collected wrist sEMG data from 43 participants over three separate days.
• Performed cross-day analyses using training and testing data from different days.
• Investigated three fusion levels: score-level, rank-level, and decision-level fusion.

Main Results:

• The score-level fusion scheme achieved a median rank-1 accuracy of 77.9% and rank-5 accuracy of 99.6%.
• These accuracies were significantly higher (p<0.001) than those obtained from single-code gestures.
• The multicode sEMG biometric framework demonstrated superior identification performance in cross-day evaluations.

Conclusions:

• The multicode sEMG biometric framework provides enhanced identification performance in realistic cross-day scenarios.
• Score-level fusion is an effective strategy for improving sEMG-based personal identification accuracy.
• sEMG biometrics, particularly using a multicode approach, present a promising spoof-resistant identification solution.