Anticipatory prediction of sit-to-stand and stand-to-sit transitions: a unified approach

Introduction Sit-to-stand ( S i TS t ) and stand-to-sit ( S t TS i ) motions, collectively referred to as STS motions, are fundamental movements for independent daily living. However, many individuals are unable to generate sufficient strength and balance during these motions, which increases the risk of fall-related accidents. Therefore, proper and timely mechanical assistance is needed to improve the quality of daily life for people with weak muscles or insufficient motor control. Methods To support the development of such assistance, we analyzed S i TS t and S t TS i ) motions performed using two different strategies by means of electromyography. Muscle synergy analysis was used to provide a compact and physiologically interpretable description of myoelectric patterns, enabling systematic comparisons of neuromuscular control across the two movement strategies, namely the momentum transfer strategy and the stabilization strategy. Based on these findings, we further proposed a deep neural network framework to predict motion states prior to motion initiation. Results The experimental results demonstrated that at least four synergy patterns were sufficient to represent these STS motions. In addition, the proposed method achieved an accuracy of 92.97 ± 0.86% with a forecasting time of 300 ms for motion state prediction, while the average temporal error remained consistently below 50 ms. Discussion These findings indicate that muscle synergy analysis can effectively characterize different STS movement strategies and that the proposed deep neural network framework can provide sufficient lead time for assistive device activation. This approach may contribute to the development of effective mechanical assistance systems for individuals with impaired muscle strength or motor control.