I have a one-dimensional matrix (that was extracted from the audio data) to process so the noise of the data could be eliminated. A paper told me that I have to use the Karhunen-Loeve Transformation (KLT). But, when I search for the fundamental of KLT, it involves eigenvectors and eigenvalues of the matrix's covariance. I have 1 x n or n x 1 matrix so the covariance is a certain value not a matrix (or 1 x 1 matrix). Then, the eigenvectors must be one. But the KLT transform is: y = eigenvector(covariance(data))*data.. It is not changing anything. Is one-dimensional matrix can be transformed by KLT?