Automatic user identification by using forearm biometrics

Electromyography (EMG) based human machine muscle interfaces hold great potential for interfacing the complexity of our body, with a multitude of electronic devices. However, the lack of compensationary methods for adapting systems from one user to another, prevents us achieving easy to use devices. This paper presents a method for enhancing EMG usability, which is based on biometrically identifying a user, so that previous training data can be automatically retrieved. This minimizes the need for small groups of people to repeatedly re-train a system over a short to medium time frame. Experiments were performed to test how EMG, circumference, as well as a combination of both, can be used as a biometric for identifying 4 users, in small group sizes of 4, 10 and 19. The results show average identification accuracies across all 11 gestures of 55.32%, 75.44% and 90.32%, for groups of 19,10 and 4 subjects respectively, while attaining the best single gesture identification accuracies of 60.04%, 82.8% and 100%.