Human intention estimation method for a new compliant rehabilitation and assistive robot

Focusing on physical human-robot interaction, biosignal feedback allows a rehabilitation and assistive robotic system possess many advantages, such as providing human intention estimations, suitable treatment evaluations, capacity of generating power assistive strategies in advance, etc. In general, electromyogram (EMG), angle, and force signals can be used to estimate the intention of the human by two kinds of human intention estimators, such as binary intention and continuous intention estimators. In this paper, we propose a hybrid estimator of human intention using a Support Vector Machine (SVM) and Linear Regression to recognize human intention and apply it to a new rehabilitation and assistive system, the coupled elastic actuation robotic system (CEBOT), which is designed to enhance the human mobility. With unique intrinsic adjustable stiffness and human intention recognition capacity, the CEBOT, possessing inherent safety, gentler treatment, suitable motion patterns, capacity of combining functional electric stimulation, etc. The design and investigations of the system are provided and discussed. The proposed method for human intention estimations is verified by experimental results.