Estimation of Cylinder Grasping Contraction Force of Forearm Muscle in Home-Based Rehabilitation Using a Stretch-Sensor Glove

Monitoring forearm muscle contraction force in home-based rehabilitation remains challenging. Electromyography (EMG), as a standard technique, is considered impractical and complex for independent use by patients at home, which poses a risk of device misattachment and inaccurate recorded data. Considering the muscle-related modality, several studies have demonstrated an excellent correlation between stretch sensors and EMG, which provides significant potential for addressing the monitoring issue at home. Additionally, due to its flexible nature, it can be attached to the finger, which facilitates the logging of the kinematic mechanisms of a finger. This study proposes a method for estimating forearm muscle contraction in a cylinder grasping environment during home-based rehabilitation using a stretch-sensor glove. This study employed support vector machine (SVM), multi-layer perceptron (MLP), and random forest (RF) to construct the estimation model. The root mean square (RMS) of the EMG signal, representing the muscle contraction force, was collected from 10 participants as the target learning for the stretch-sensor glove. This study constructed an experimental design based on a home-based therapy protocol known as the graded repetitive arm supplementary program (GRASP). Six cylinders with varying diameters and weights were employed as the grasping object. The results demonstrated that the RF model achieved the lowest root mean square error (RMSE) score, which differed significantly from the SVM and MLP models. The time series waveform comparison revealed that the RF model yields a similar estimation output to the ground truth, which incorporates the contraction–relaxation phases and the muscle’s contraction force. Additionally, despite the subjectivity of the participants’ grasping power, the RF model could produce similar trends in the muscle contraction forces of several participants. Utilizing a stretch-sensor glove, the proposed method demonstrated great potential as an alternative modality for monitoring forearm muscle contraction force, thereby improving the practicality for patients to self-implement home-based rehabilitation.