Inductive cloning of a state model for normal gait

The technique of rule-based inductive learning was used to ‘clone’ the control rules used for muscle activation from kinematic variables measured during normal walking. The activation patterns (or desired output) consisted of the EMG signal of the semitendinosus muscle, determined from surface electrodes. The attributes (or input) consisted of kinematic variables. It was possible to reconstruct a quantised form of the muscular activation. The induced rules used were both explicit and comprehensible.