A perceptually based audio signal model with application to scalable audio compression