A comparative study of constrained and unconstrained approaches for segmentation of speech signal