PCA LDA

Hello, I am doin a project on face detection+recognition. Face detection part is working properly. I hav done face recognition using PCA but this method is highly undependable, so i planned to use LDA once we get the weights using PCA. However LDA algo gives me a eigenvector of size m*m where m=number of classes(in my case faces). However for recognition testing the vector will be of size M*1. where M=no of eigen vectors using PCA and i hav to project it on the LDA space but the dimensions arent matching. so ther is an error. Please someone help me out. I am stuck at this point.

[Ilya]

If you show your code, it would help understand what you are asking.

MATLAB implementations of LDA, either function CLASSIFY or the new class ClassificationDiscriminant, compute m*(m-1)/2 sets of linear coefficients for m classes. I don't know what you mean by "eigenvector of size m*m".

Perhaps you want to project your data onto the principal components and then apply LDA to the projected data. It is hard to be certain based on your description that that's what you want.

Answer 1

[Matlaber]

Why deleted?