Muscle Activation Visualization System Using Adaptive Assessment and Forces-EMG Mapping

Abstract

VR-based serious games are obtained without details about real-time guide and feedback in the rehabilitation after stroke, leading to undesirable recovery outcomes. This study investigated the feasibility of real-time visualization in muscle state feedback by sEMG. Then we explored the application in movement guide and diagnosis. We provided a force-sEMG mapping approach based on body weight to visualize the implicit bioinformation. And 10 healthy subjects participated in an experiment that the K-means cluster algorithm and the support vector regression are employed to filter the unexpected data and adjust the visualization parameter for each subject dynamically. The verification experiment demonstrates that force and sEMG can be mapped as a data pair by support vector regression and normalized by the low-cost calibration in 6–8 times reparative actions. We define the predictive accuracy as the ratio of the predicted to the practical tasks. And mean absolute value is the most suitable index to compete for most data scales which can provide a predictive accuracy of 89.84% ± 5.28% in biceps and 84.26% ± 6.44% in triceps. We present the motivation improvement for patients and supervision in online and offline application scenarios for therapists. This system can be applied in precision and telemedicine to improve the efficacy of rehabilitation by objective data and intuitive expression through visualization.

Keywords

• K-means cluster
• rehabilitation
• surface electromyography
• stroke
• SVR
• visualization