Waveform approximating residual audio coding with perceptual pre- and post-filtering

We investigate waveform approximating residual coding for a sinusoidal parametric audio coder at low bit rates. The residual coding is based on the well-known pre- and post-filtering method with lossless coding which features perceptual weighting for short time segments. We compare the incurred perceptual distortion from joint quantization of the residual and the sinusoids for different bit rates. In addition to that, we develop a transform coding scheme for the coefficients in the pre- and post-filters which must be send as side information between the encoder and decoder. Our investigations show that the combination of the sinusoidal subcoder and the pre- and post-filtering entails an overall lower perceptual distortion for low as well as high bit rates. Also, the developed transform coding scheme enables efficient coding of the side information at a low bit rate.