FeatureSet method: matmul

Syntax: <feature set> matmul <new feature> <source feature> <FMatrix> [-cut <dim>]

Example: fs matmul LDA MEL matrixA -cut 10

Each frame of the <source feature> is multiplied with a <FMatrix>. So if the <FMatrix> has m rows and n columns each frame of the <source feature> must have n coefficents and the <new feature> will have m coefficients. With <dim> < m only the first <dim> coefficients can be selected for the <new feature>. In the example each frame of the feature MEL is multiplied with the first 10 rows of matrixA and the result is stored framewise in a new feature LDA. LDA and MEL will have the same number of frames.

Optional flags:

-cut <dim>

Select the first <dim> coefficients. In other words only the first <dim> rows of the FMatrix will be taken for the multipication.