EEG signal classification during listening to native and foreign languages songs

This paper designs an experiment to analyze different EEG patterns while subjects are listening to different language songs. In the process of experiment, the subjects listen to multi-section songs. Every two songs have the same rhythm and only the lyrics are different, one in Chinese and the other in Japanese. The songs are sung by one singer and the Chinese subject dont know Japanese at all. At the same time we collect the EEG signals which are supposed to have very subtle difference corresponding to two kinds of songs. Then we use common spatial pattern algorithm to extract features and define an average energy function to represent them. After that we use support vector machine to learn and classify the EEG data. We find that the difference pattern mainly lay in low spectral band (0-0.5 Hz), and concentrate on the left frontal area of the cortical. We achieve the highest classification accuracy of 97.30% and an average classification accuracy of 87.15%.