Code for Common spatial pattern

Hello,

I am trying to implement Common Spatial pattern technique to extract some features from Deap dataset for classification. Can somebody share his code of CSP with me as I am not getting very good variance separation between classes. I am trying it with the file available in the link http://in.mathworks.com/matlabcentral/fileexchange/22915-common-spatial-patterns.

May be I am doing some mistake in extracting the features from the filtered signal. It will be helpful for me if someone show me the way to extract features correctly.

Any comments or suggestions....

Upasana

[xiao]

If you have finished the classification algorithm of DEAP dataset based on csp, I feel a little confused. I hope you can share your paper and code for me to learn. Thank you very much.

[behtam adeli]

The code that has been linked just produces the Projection matrix as it has been mentioned at the end of the code (%Resulting Projection Matrix-these are the spatial filter coefficients).

You need to multiply this matrix with you original data and then extract your feature. A famous feature is the log of variances like this:

Z = Projection{class}([1,end],:)*orginialData'; 
feature =log([var(Z(1,:)),var(Z(2,:))]./(var(Z(1,:))+var(Z(2,:))));

I hope this helps someone.