Surface Electromyography-Based Personal Identification Using Automated Machine Learning: A Promising Approach for Biometric Authentication

Le résumé fourni par la source

Traditional biometric methods, such as facial recognition, fingerprints, and iris scans, are widely used but are susceptible to spoofing. Electromyography (EMG) signals, which reflect the electrical activity of muscles during activation, offer a promising alternative for biometric authentication due to their uniqueness, stability, and difficulty to forge. This study aimed to explore the potential of EMG signals as a reliable biometric feature for personal identification and to evaluate the performance of different machine learning models in this context. EMG signals were collected from ten volunteers using eight surface EMG electrodes placed on the forearm muscles during ten different hand movements. The dataset was analyzed using logistic regression, decision trees, random forests, and automated machine learning (AutoML). Model performance was assessed using accuracy, precision, recall, and F1-score. The AutoML model demonstrated superior performance, achieving an accuracy of 0.80, outperforming logistic regression (0.31), decision trees (0.36), and random forests (0.77). The random forest model also showed strong results, particularly in precision and recall for certain individuals. However, performance variability was observed across different individuals, indicating potential influences of individual physiological differences on model effectiveness. EMG signals hold significant potential as a biometric modality for personal identification. Automated machine learning approaches, especially AutoML, are highly effective in leveraging the complex patterns within EMG data, offering a robust solution for biometric authentication. Future research should focus on enhancing model adaptability for individuals with lower recognition accuracy and exploring multimodal biometric systems that integrate EMG with other biometric signals to further improve identification performance and reliability.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.