Modelling the relationship between human perception and sound quality parameters using LS-SVMs