BCI competition 2005: Data set I

Group information:

Institute of Biomedical Engineering of Xi’an Jiaotong University

Xi’an, 710049, P. R. China

Team Member: Xiaomei Pei, Guangyu Bin

pei@zy165.com

Supervisor: Prof. Chongxun Zheng.

1. Description of our method

Firstly, according to Fisher ratio, the 4 channels with most reactive power features are selected.

Secondly, by EEG signal of 4 channels within 3 sec, 2-order MVAR (multi-variate AR) model is established and the MVAR coefficients are extracted as feature vectors for further classification.

Finally, by Fisher Discriminant Analysis, the two EEG patterns could be discriminated.

2. Format of the data

The result is saved in a Matlab-file format. res.mat is the classification output with labels -1, 1 of the test data.