PURPOSE:To control the fingers of a robot hand according to the operating recognition of the fingers of a hand depending to the skin surface electromuscle by recogniting the electromuscle of a robot hand operator having skin surface electrodes by use of a neural circuit net with a learning function which is taught and learned per kinds of hand actions, thus controlling the robot hand. CONSTITUTION:The electromuscle of a normal operator or forearm cut operator due to the hand actions of the operator' intension is detected by skin surface electrodes 1a, 1b equipped in the vicinity of the muscle related to the hand actions. The detected electromuscle signals are subjected to Fourier transformation in a transformation part 9 and the frequency analytic pattern and the pattern impossible in a linear separation are taught and learned in plural times by a neural circuit net capable of separation per kinds of respective patterns. Thus, the patterns corresponding to the hand actions are recognited and discriminated by a recognition part 13, thereby controlling a robot hand 19 by a control part 17. Therefore it is possible to control the fiber actions of the robot hand 19 according to the operating recognition of the fingers of a hand depending to the skin surface electromuscle.
展开▼