EEG signals classification of motor imagery based on multi-feature description

Electroencephalogram (EEG) based Brain Computer Interface (BCI) provides a new communication and control channel for people with severe motor disabilities. Motor imagery analysis is one of the widely used methods in the BCI field. However, these motor imagery signals are very noisy and strongly depended on subjects. Therefore, more powerful classification methods are needed. In this paper, a novel classification method is proposed based on feature selection. In training mode, all features for all channels are calculated first. Then, a Genetic Algorithm (GA) is used to search for the best feature set. Once feature set are determined, in testing mode, only those features selected are calculated and used to make the final classification. The corresponding experiment results show that, GA has the ability of finding the most useful features and with the feature selection, the final classification accuracy is improved clearly. All the experiment results demonstrate the feasibility and effectiveness of the proposed method.