Control of FES in paraplegia: modeling voluntary arm forces

Skilled behavior is difficult or impossible to articulate explicitly by the performers. Likewise biomechanical models of skilled motor actions are often limited by the lack of knowledge of the underlying mechanisms. A 'behavioral cloning' technique is described, based on a trained artificial neural network (ANN), that precisely mimics an individual's learned skill. In this paper the motor skill considered is that of paraplegics using their upper limbs whilst standing-up with FES. In a group of eight paraplegics with complete spinal injuries, it was possible to develop clones that followed closely the observed behavior of the subjects. Each subject used a unique and consistent voluntary control strategy. Subjects with more experience in using FES were more consistent in the use of their arms from trial to trial. Comparison of the clones revealed features suggestive of some common underlying voluntary control strategies.