Virtual pattern classification of upper limbs motion using artificial neural networks

Abstract

Virtual reality technology is common used to entertain people as movies or games. At present, this technology applies to medical field for training surgeon on operating simulation or patients with either neurological disease or psychiatric disorder. The study focused on the algorithm of pattern classification. The artificial neural network was considered to achieve this classification. The multilayer perceptron with four input nodes, thirty nodes in hidden layer and five output nodes were designed for this classification algorithm. The virtual reality showed the animator who acted as the trainer. The movement of trainer was used to be the supervised data of the neural network. The users moved their arms along with the animator and recorded the motion. These data were the testing data set of network. The results showed that the neural network could classify all motion patterns. It was difficult to classify the patterns in the same side Pattern 5 was correctly classified by this neural network model. © 2013 IEEE.