EMG-based body–machine interface for targeted trunk muscle activation

Deficits in trunk control, commonly observed in individuals with neurological conditions, can significantly impair balance, posture, and functional movements. Body–machine interfaces (BoMIs) are promising tools for trunk rehabilitation, as they can provide real-time feedback on user movements and muscle activity, allowing for continuous monitoring and guidance during motor control training. However, research on BoMIs for trunk rehabilitation is limited, and current methods often lack precision in addressing trunk muscle deficits. This work introduces a BoMI that combines trunk electromyography (EMG) and motion data to selectively modulate trunk muscle activity during motor control tasks. The system utilizes machine learning to generate personalized trunk motion trajectories based on predefined EMG profiles. Each trajectory is displayed on a screen as a moving target, which users must follow by controlling the BoMI with their trunk movements. We hypothesize that by visually guiding users to track these generated trajectories, the BoMI could evoke the EMG patterns implicitly encoded within them. Tested with neurotypical individuals, the BoMI effectively elicited the desired trunk EMG profiles, achieving a mean similarity index of 0.82 ± 0.13, a correlation coefficient of 0.95 ± 0.03, and minimal timing mismatches. These results support the feasibility of using an EMG-based BoMI for precise trunk muscle training, which could potentially assist therapists in more efficiently monitoring and adjusting patients’ muscle engagement during interventions. Future work will focus on developing a control framework to dynamically adapt task difficulty to users’ needs, expanding the approach to include additional trunk muscles, and evaluating its translation to individuals with trunk muscle impairments.