Serious gaming to generate separated and consistent EMG patterns in pattern-recognition prosthesis control

Pattern-Recognition (PR) control of upper-limb prosthetics has shown inconsistent results outside lab settings, which might be due to the inadequacy of users' electromyogram (EMG) patterns. To improve the separability and consistency of their EMG, users can receive training. Conventional training uses an internal focus of attention as prosthesis users focus on the muscle contractions of their (phantom) hand together with explicit learning processes facilitated by a coach guiding the user. in this study we investigated if an alternative training paradigm using an external focus of attention exploiting implicit learning processes based on serious gaming without a coach could lead to more separable and consistent EMG. Able-bodied participants (N = 25; mean age 22 years, 13 females) were recruited and followed conventional or game training for five days. In conventional training, participants performed the Motion Test thrice daily and received coaching on how to adapt their muscle contractions. In game training, participants controlled an avatar using a direct mapping from electrode to avatar direction. The participants utilized implicit learning processes, by exploring which muscle contractions made the avatar go in which directions. Performance in both groups was evaluated by using the Motion Test in a pm/post-test design. Training resulted in improved performance, with no differences between training paradigms. Participants who followed game training showed 51% more separated EMG patterns. EMG pattern consistency did not change over training. It was concluded that serious game training using an external focus of attention and implicit learning can be considered as a viable alternative to conventional training.