Predicting hand orientation in reach-to-grasp tasks using neural activities from primary motor cortex

Hand orientation is an important control parameter during reach-to-grasp task. In this paper, we presented a study for predicting hand orientation of non-human primate by decoding neural activities from primary motor cortex (M1). A non-human primate subject was guided to do reaching and grasping tasks meanwhile neural activities were acquired by chronically implanted microelectrode arrays. A Support Vector Machines (SVMs) classifier has been trained for predicting three different hand orientations using these M1 neural activities. Different number of neurons were selected and analyzed; the classifying accuracy was 94.1% with 2 neurons and was 100% with 8 neurons. Data from highly event related neuron units contribute a lot to the accuracy of hand orientation prediction. These results indicate that three different hand orientations can be predicted accurately and effectively before the actual movements occurring with a small number of related neurons in M1.