Fig. 10
From: The concepts of muscle activity generation driven by upper limb kinematics
The recurrent neural network (RNN) is separately trained on several parameters: angular position (ang), angular velocity (vel), and angular acceleration (acc) from each joint as well as on the end-effector (EEF) position and orientation of the hand and EEF+ with additional velocity and acceleration of the EEF. The RNN is tested on a new motion from the separate subject a-c and evaluated by zero-line score Zs. The results are presented as an average over all channels