Abstract: This paper presents a signal model for scalable perceptual audio coding consisting of Sines + Noise (SN) representations. The paper essentially presents a fundamental enhancement to the sinusoidal modeling component. The enhancement involves an audio signal scheme based on carrying out overlap-add sinusoidal modeling at three successive time scales, large, medium, and small. The sinusoidal modeling is done in an analysis-by-synthesis overlap-add manner across the three scales by using a psychoacoustically based matching pursuits. The sinusoidal modeling residual at the first scale is passed to a couple of smaller scales to allow for modeling of various signal features at appropriate resolutions. This approach greatly helps to correct the pre-echo inherent in the sinusoidal model. The new scheme gives an improved perceptual audio quality compared to our previous work while using the same number of sinusoids.

Key words: mulfiresolution sinusoidal modeling, parametric audio coding, low-rate audio coding, signal modifications.