[Eeglablist] a question about csp code ,please help me

ahmad hasanpour a.ahp89 at gmail.com
Thu Sep 15 11:40:10 PDT 2011


please help me my friends,i see  csp code in bci lab ,let me describe for
you understand i get it; first create k fold with 10,and then take set of
trials different for obtain good spatial filter?(i cant understand this
section please describe it)

after u obtain good spatial filter u  apply it on test trails.

i do this steps on datasetIIIa but i cant get good result(2 classes).

cov1=cov(signals(1:400,:));        %Class 1
cov2=cov(signals(803:1203,:));    %Class2

[W D]=eig(cov1,cov1+cov2);

filter=W(:,[1:3 56:59]);                % M first and end W

sig1=signals(1:400,:)*filter;

sig2=signals(803:1203,:)*filter;
 log(var(sig1))
 -4.2778   -3.6758   -2.9707   -0.1514   -0.1446   -0.1144   -0.0163

log(var(sig2))
  -0.0140   -0.0257   -0.0526   -1.9627   -2.0050   -2.2249   -4.1225

sig3=signals(402:802,:)*filter;   *%Class1  for Test*
log(var(sig3))
 0.0388   -0.9152    0.1855   -0.2021   -0.0560   -0.8266   -1.8903

i dont have good result;why?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110915/13741aeb/attachment.html>


More information about the eeglablist mailing list