1.3 Spatial audio coding

Thus, the trend towards high-quality, multi-channel audio for solid-state and mobile applications imposes several challenges on audio compression algorithms. New developments in this field should aim at unsurpassed compression efficiency, backward compatibility with existing systems, have a low complexity, and preferably support additional capabilities to optimize playback on mobile devices. To meet these challenges, the field of spatial audio coding has developed rapidly during the last 5 years. Spatial audio coding (SAC), also referred to as binaural cue coding (BCC), breaks with the traditional view that the amount of information that has to be transmitted grows linearly with the number of audio channels. Instead, spatial audio coders, or BCC coders, represent two or more audio channels by a certain down-mix of these audio channels, accompanied by additional information (spatial parameters or binaural cues) that describe the loss of spatial information caused by the down-mix process.

Conventional coders are based on waveform representations attempting to minimize the error induced by the lossy coding process using a certain (perceptual) error measure. Such perceptual audio coders, for example MP3, weight the error such that it is largely masked, i.e. not audible. In technical terms, it is said that ‘perceptual irrelevancies’ present in the audio signals are exploited to reduce the amount of information. The errors that are introduced result from removal of those signal components that are perceptually irrelevant.

Spatial audio coding, on the other hand, represents a multi-channel audio signal as a down-mix (which is coded with a conventional audio coder) and the before mentioned spatial parameters. For decoding, the down-mix is ‘expanded’ to the original number of audio channels by restoring the inter-channel cues which are relevant for the auditory system to perceive the correct auditory spatial image. Thus, instead of achieving compression gain by removal of irrelevant information, spatial audio coding employs modeling of perceptually relevant information only. As a result, the bitrate is significantly lower than that of conventional audio coders because the spatial parameters contain much less information than the (compressed) waveforms of the original audio channels. As will also be explained in this book, the representation of a multi-channel audio signal as a down-mix plus spatial parameters not only provides a significant compression gain, it also enables new functionality such as efficient binaural rendering, re-rendering of multi-channel signals on different reproduction systems, forward and backward format conversion, and may provide means for interactivity, where end-users can modify various properties of individual objects within a single audio stream.