Precise Recognition of Complicated hand Operations Based on EEG and Master-Slave Neural Network

As a good classifier, BP neural network has been applied in many engineering research questions. However, because of some inherent shortages, especially chaotic behaviors in the network learning, it is very difficult or impossible to apply the artificial neural network into the precise recognition of the complicated hand operations based on electroencephalography (simply denoted as EEG). Based on good properties of the Hopfield neural network, a new master-slave neural network model (simply denoted as MSNN) is presented in this paper firstly, whose master network is two Hopfield networks, and the other slave network is a BP network, respectively. After its structure had been innovatively designed, the training algorithm of the MSNN was simply discussed. And then, a two- channel EEG measurement system was set up, and the feature of the related EEG signals extracted. At last, some complicated hand operations are respectively recognized by using the MSNN and BP neural network. The comparable analysis results showed that the MSNN had a better asymptotic convergence rate and a higher mapping precision, so that it gave higher recognition possibilities than the BP network did, whose recognition possibility was improved from 50%, 40%, 40%, 40%, 50%, 40%, and 50% to 70%, 60%, 70%, 70%, 80%, 60%, and 80% for grasping, relaxation, dynamic grasping, dynamic loosing, grasping a small bar, grasping a hard paper, and grasping a baseball of the seven complicated hand operations, respectively.