Should We Look at Curvature or Velocity to Extract a Motor Program?

Experimental studies led by Lashley and Raibert in the early phase of human movement science highlighted the phenomenon of motor equivalence, according to which complex movements are represented in the brain abstractly, in a way that is independent of the effector used for the execution of the movement. This abstract representation is known as motor program and it defines the temporal sequence of target points the effector has to move towards to accomplish the desired movement. We present and compare two algorithms for the extraction of motor programs from handwriting samples. One algorithm considers that lognormal velocity profiles are an invariant characteristic of reaching movements and it identifies the position of the target points by analysing the velocity profile of samples. The other algorithm seeks target points by identifying the trajectory points corresponding to maximum curvature variations because experimental studies have shown that the activity of the primary motor cortex encodes the direction of the movement. We have compared the performance of the two algorithms in terms of the number of virtual target points extracted by handwriting samples generated by 32 subjects with their dominant and non-dominant hands. The results have shown that the two algorithms show a similar performance over 55% of samples but the extraction of motor programs by analysing the curvature variations is more robust to noise and unmodeled motor variability