Abstract

The online control mechanisms of human goal-directed movements include two functions: predicting sensory feedback resulting from motor commands (forward transformation) and generating motor commands based on sensory inputs (inverse transformation). Infants acquire these transformations without implicit instructions from caregivers. In this paper, a neural network model is proposed that learns the forward and inverse transformations in reaching movement by observing the randomly moving hand. The forward pathway of the model is a Jordan network that is input with motor commands, and that is trained to output a visual hand position. The inverse pathway has the input of the visual hand position and a connection from the hidden layer of the forward pathway. It is trained to output the motor command, which makes the hand move to the input hand position. The network learned correct transformations, which suggests that continuous observation of the hand is the basis for motor development.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call