Abstract
Electroencephalograph (EEG) recorded during left and right hand motor imagery can be used to move a cursor to a target on a computer screen (such systems are called a brain computer interface: BCI). A BCI has been studied for one of rehabilitation programs intensively. However, the spatial resolution of EEG is inferior to other acquisition methods. Focusing attention on new information of EEG is required. In this paper, we proposed the Pulse Complex Model (PCM) as a new pattern recognition model to extract features from EEG concerning with motor imagery. In applying to the model, the parameters were estimated by using a Genetic Algorithm (GA). Then a discrimination rule based on a Support Vector Machine (SVM) was constructed. From the results, the best preprocessing and the optimal order of the model were estimated. On the basis of these results, some types of discriminant analyses were conducted. According to the results, a relation of approximation errors to the feature of motor imagery was revealed.
