(Semi)parametric Audio Coding

Modern perceptual audio codecs that target low bitrates are usually equipped with a set of coding tools that enable an efficient parametric (and semi-parametric) representation of the signal, i.e. they describe the signal by a sparse set of parameters rather than a time or frequency representation. As such, these parametric coding tools are typically not waveform-preserving, i.e. even with an infinite bit budget, the reproduced waveform will not converge to the original waveform.

Typically, such tools can be used to achieve good quality at very low bitrates (rather than full perceptual transparency at high bitrates). Due to their non-waveform-preserving nature, they can – if used improperly – introduce characteristic artifacts that are demonstrated within the following chapters of this section:

• Perceptual Noise Substitution
• Bandwidth extension
• Parametric stereo/multi-channel