EEG Based Brain Computer Interface On Hearing

Language is human's tool for everyday communication. Based on the hearingperspective, we use EEG based method to study the mechanism of language processingin brain. We design the experiment which uses Chinese and Japanese songs as stimulimaterials and totally get 23 EEG data. We use ICA, linear filter to preprocess signal,CSP to extract the feature, and classifier to train the data. The result indicate thatSVM classifier has better generalization ability than MLP. SVM get 82.41% averageaccuracy for eight subjects. This paper firstly uses the fuzzy integral method in EEGanalysis and the result shows that fuzzy integral has better classifying accuracy thanthe method which directly splices different spectral feature of the signal. The fuzzymethod gets the average accuracy of 86.49% for eight subjects. Finally, we use theunsupervised learning method to represent the feature of two classes. The result showsthat language processing area of brain is located in the left temporal and inferior frontalgyrus area and is correlated with the SCP, Beta and Gamma signal.The contribution of this paper are:Firstly, we designed the hearing-evoked language task of EEG experiment. Weget lots of different songs as stimuli signal and collect some continuous EEG;Secondly, we use common spatial pattern, MLP and SVM to analyze the signal,especially we introduced the fuzzy integral algorithm in EEG analyzing;Finally, we use the unsupervised method to get the feature of the two classessignal and find the language related brain area.