Uncorrelated optimal discriminant vectors based on generalized singular value decomposition

In this paper, we put forward an uncorrelated optimal discriminant vector algorithm based on generalized singular value decomposition for feature extraction. The algorithm maximizes the Fisher criterion by employing the theory of generalized singular value decomposition, and obtains projection vector with the constraint of statistical uncorrelated theory. Especially, our method provides a mathematical method to contribute to understanding the problem of singular by employing the generalized singular value decomposition technique. Experimental results show that our method reduces the computational complexity and gets better performance.