Hybrid Model-Based Classification of the Action for Brain-Computer Interfaces

Artificial Intelligence has made tremendous progress in industry in terms of problem solving pattern recognition. Mirror neuron systems (MNS), a new branch in intention recognition, has been successful in human robot interface, but with some limitations. First, it is a cognitive function in relation to the basic research limited. Second, it lacks an experimental paradigm. Therefore MNS requires a firm mathematical modeling. If we design engineering modeling based on mathematical, we will be able to apply mirror neuron system to brain-computer interface. This paper proposes a hybrid model-based classification of the action for brain-computer interface, a combination of Hidden Markov Model and Gaussian Mixture Model. Both models are possible to collect specific information. This hybrid model has been compared with Hidden Markov Model-based classification. The recognition rates achieved by Hidden Markov Model were 76.62% and the proposed model showed 84.38%.