SOUND TEXTURE MODELLING WITH LINEAR PREDICTION IN BOTH TIME AND FREQUENCY DOMAINS