GRU-Based Prediction of Muscle Activation Using IMU Data: Integrating VR for Rehabilitation Assessment

In this paper, we propose a novel approach using IMU data to predict muscle activation, diverging from the conventional method of employing sEMG data for continuous angle estimation. We opt for Gated Recurrent Units (GRUs) over Long Short-Term Memory (LSTM) due to their simplified and efficient structure, achieving an accuracy of 0.7562 in predicting muscle activation. Additionally, we utilize VR to visualize the color changes in muscles, enhancing the effectiveness of rehabilitation assessments.