“…While the classifier is used to understand which hand or grasp posture is desired (e.g., fully stretched hand, power grasp, pointing index), the regressor predicts the force required in the case of a grasping posture. The approaches presented so far in the literature use various techniques for preprocessing the data [2], [3], classifying the postures [3]- [5] and predicting the associated force [3], [6], [7]; despite the differences, all these studies share a common validation procedure, in which experiments are conducted on a proprietary database containing between five and ten intact subjects and amputees, which are executing up to ten different grasp actions, enacting static hand postures or moving their fingers and wrist. (We will from now on collectively denote these actions movements of interest; see section I-A for a more detailed review).…”