Fig. 1

a Typical profiles of velocities of the elbow and hand aperture in abled-bodied subject [21, 45, 46] compared to that generated with a traditional prosthetic device, as presented in [25, 26]. During reaching, the aperture of the human hand (solid green line) changes in coordination with the extension of the arm (dashed blue line). In contrast, the prosthetic hand (dash-dotted red line) begins its motion later in the reach-to-grasp cycle, once the elbow is fully extended. In our approach, we separate the reach-to-grasp motion into three phases (denoted by dashed vertical lines) according to the angular acceleration of the elbow joint a _el . We distinguish between acceleration, deceleration and rest phases. We present that a pattern recognition system, trained including the reaching motion, could gain efficient prediction confidence early in the reaching motion and, thus, activate faster a prosthetic device. b The selected five grasp types used in our classification, following the names and using figures from the taxonomy of [47]. c Experimental set-up for training the system with amputee subjects in data recordings. EMG-information from the amputated arm are recorded while the subject performs the reach and grasp motion with his/her intact arm