Abstract:

BACKGROUND: Reaction velocity for brain imagination is a major criterion in measuring quality of brain computer interface (BCI) system. Therefore, electroencephalogram (EGG) analysis, especially the feature extraction screening, is very important. Reduction the number of features is an important way to improve the speed.
OBJECTIVE: To screen EGG features using reduction algorithm, and to decrease the number of EGG features. 
METHODS: Firstly, by various analytical methods, the EGG features were extracted and classified; then, discreting the continuous EGG to establish an EGG information table; At last, reduction theory was used to reduce EGG features, and to sort the data according to reduced attributes, the accuracy of sorting was validated.
RESULTS AND CONCLUSION: Using reduction algorithm and feature marking, discreting the continuous EGG to establish an EGG information table, and then choose the features from the discrete data. The results show that classification accuracy has not been reduced but the number of features was reduced. However, this method can be used marking features in two sorts, how to marking features of multi-sorts and to perform data reduction is a key point in further study.