mcsech
Newbie level 6
Hi guys.
I'm working on Voice Conversation project. Briefly, I have two trained data belongs two different people. With a Codec I'm trying to analyise trained data.
Now,I've trouble with this problem.
i.e X=[] is person K, and Y=[] is person L. X and Y are 2D matrices.
With the canoncorr function of matlab I want to a relation between two matrices. The function description says:
[A,B]=canoncorr(X,Y) computes something like this:
A.X=B.Y;
Then I think H=inv(B)*A, and Z=H*X' (X' is transpose of X).
Theoretically, the transpose of Z matrix should similar to Y. But something is wrong.
Could anyone help me? You can advise a different method to find correlated Z matrix.
I'm working on Voice Conversation project. Briefly, I have two trained data belongs two different people. With a Codec I'm trying to analyise trained data.
Now,I've trouble with this problem.
i.e X=[] is person K, and Y=[] is person L. X and Y are 2D matrices.
With the canoncorr function of matlab I want to a relation between two matrices. The function description says:
[A,B]=canoncorr(X,Y) computes something like this:
A.X=B.Y;
Then I think H=inv(B)*A, and Z=H*X' (X' is transpose of X).
Theoretically, the transpose of Z matrix should similar to Y. But something is wrong.
Could anyone help me? You can advise a different method to find correlated Z matrix.
Last edited: