I have two signals, I've done a Principal Component Analysis and separated my signals, I get one with noise and one with signal. Before PCA and seperation After PCA and seperation.

X = [Data1,Data2]';
Q = X*X';
[W,D] = eig(X*X',Q);     
S = W'*X; 

The problem is PCA normalizes my signal (the scales are vastly different). I would like to preserve the scale accurately, do I choose different eigenvectors?

  • $\begingroup$ Could you please show all the details of the process that lead to the diagrams? $\endgroup$ – Jazzmaniac Aug 2 '16 at 10:44
  • $\begingroup$ After I find the eigenvectors and apply them to the data, I do this:figure;plot(S) S is a 2x7000 matrix $\endgroup$ – Voltage Spike Nov 28 '16 at 19:02
  • $\begingroup$ You could do the normalization yourself for each signal and then after PCA transform the result back. Therefore you will only have to record both mean and std for each dimension. whereas z = (x-mean)/std, you can just as well go back as x = z*std + mean $\endgroup$ – Nikolas Rieble Dec 12 '16 at 13:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.