摘要 : The paper reviews the classification approaches of non-invasive P300 and hybrid brain-computer interface (BCI) spellers with threshold over 85% of classification accuracy and high speed (the best speed among reviewed is about 131 ... 展开 The paper reviews the classification approaches of non-invasive P300 and hybrid brain-computer interface (BCI) spellers with threshold over 85% of classification accuracy and high speed (the best speed among reviewed is about 131 bit per minute). The P300 spellers give patients with motor disorders the opportunity to communicate using, for instance, EEG signals. The most number of P300 spellers work with raw signal and linear classification instruments. The reviewed data could be used for elaboration of own cognitive interface not only for communication, but also for cognitive trainings and even creation. Summarizing the data we made an attempt to produce own BCI system using traditional P300 speller design, xDawn signal preprocessing and comparative linear discriminant analysis (LDA), support vector machine (SVM) and linear regression (LR) classification approaches, that demonstrated the best accuracy up to 97% for target vs non-target with LR. The tested prototype demonstrated information transfer rate (ITR) around 23 bits/min (～5 symbols per minute), that have to be improved. 收起