The identification and classification of EEG pattern features in brain-computer interface (BCI) were proposed from the angle of the intelligent processing and the uncertainty. For the uncertainty problem of the existence of EEG, two aspects of EEG processing, feature extraction and classification, were analyzed. Furthermore, we put forward the methods to solve the problem. With P300 component as an example, the channel selection, filtering and time window selection were used for feature extraction. Then the Bayes linear discriminant analysis method was used for pattern classification. Finally, the P300 data sets of the BCI competition III were used for data analysis. By comparing the classification accuracy rate of three different methods, the results demonstrated the effectiveness of our method.