I\'m currently making a program for speech recognition. In the step of codebook generation using the LBG (Linde-Buzo-Gray) algorithm, I\'ve read that the splitting factor epsilo