Data set V mental imagery, multi-class

Group information:

Center of Neuroinformatics

School of Life Science and Technology

University of Electronic Science and Technology of China (UESTC)

Chengdu, 610054, P. R. China

Supervisor: Dezhong Yao (E-Mail: dyao@uestc.edu.cn)

Team Member: Xiang Liao, Yu Yin

 

Description of our analysis method

Classification with the precomputed samples

Normalize, and classification by Support Vector Machines, used cross-validation

 

Result

 

X

 
 

 

 

 


 

 

 

 

 

 

 


Processing Flow:

(1)   Averaged by every 8 input vector of precomputed features;

(2) Plot r2 values( the proportion of the variance of the power spectral density(PSD) values accounted for by the label information) of the input vector, and select good discriminative component of the input vector as input feature vector for classification;

(3) Normalize the input feature vector to an interval of [0 1];

(4) To infer the symbols from the unlabeled test set, the Support Vector Machines (SVM) was trained on the whole training set to obtain the optimal parameter values ( regularization parameter C and the bandwidth σ of the Gaussian kernel) by fivefold cross-validation;

(5) Test set classification using the trained SVM classifier to infer the label from the test set.