Introduction

The popularity of multi-channel audio is increasing rapidly. Traditionally, multi-channel audio was driven by the movie industry. Recently, however, the music industry is shifting from stereo to multi-channel audio production as well. Content can be bought on DVD or SA-CD, and a large percentage of consumers are already equipped with multi-channel audio playback capabilities.

Despite the increasing popularity of multi-channel content, transmission of audio is in most cases still in conventional stereo format. The fact that broadcasters and service providers have not yet upgraded to multi-channel audio formats is (in part) due to (1) a lack of efficient multi-channel compression algorithms and (2) backwards-compatibility issues.

Multi-channel compression

One of the problems one encouters when dealing with multi-channel audio compression is that some of the popular audio compression algorithms (such as MPEG-1 layer 3, or mp3) simply do not support any audio channel format with more than 2 channels. Other algorithms (such as standardized in MPEG-2) do support multi-channel audio. However, the transmission bit rates increase approximately linearly with the number of audio channels. The corresponding increase in transmission bandwidth is in many cases problematic or not feasible at all.

Backwards compatibility

In important aspect of upgrading a stereo broadcast service to multi-channel audio is its compatibility with the older infrastructure. Imagine a radio station that wants to upgrade from a stereo to a 5.1 format. If this upgrade would mean that all consumers that only have a stereo compatible receiving and playback system cannot receive their favorite radio station anymore, the upgrade would most probably not take place at all.

Spatial Audio Coding / MPEG Surround

Recently, a multi-channel variant of Spatial Audio Coding has been finalized and is termed 'MPEG Surround'. MPEG Surround solves both the transmission bandwidth and the backwards-compatibility issue. In essence, it is an extension of the Spatial Audio Coding (SAC) methods as employed in Parametric Stereo but extended with multi-channel capabilities. This concept is outlined in Figure 1.

A multi-channel input signal (for example in a 5.1 format) is processed by a down mix and parameter extraction stage. The resulting down mix has a mono, stereo, or matrixed-surround compatible stereo format. This down mix can be encoded with any conventional mono or stereo compression method (for example using AAC, but other coders can be used as well). All perceptually relevant information that is lost by the down-mix process is captured by means of spatial 'parameters'. These parameters are encoded and combined with the bit stream from the conventional coder to form the total output bit stream. In most cases, the parameters can be 'hidden' in a so-called ancillary data part of the conventional audio coder bit stream. This means that conventional audio decoders will ignore this additional information layer and hence full backwards compatibility is assured.

The corresponding decoder performs the reverse process (see Figure 2). First, the conventional (mono or stereo) bit stream and parameter bit stream are separated by a demultiplexer. Subsequently, the conventional audio decoder reconstructs the mono or stereo down mix. Finally, the down mix is converted to multi-channel audio by a spatial synthesis stage which is controlled by the decoded parameters.

Spatial parameters

As mentioned above, the parameters enable a full multi-channel audio reconstruction from a mono or stereo down mix. The parameters comprise both perceptually relevant properties ('binaural cues') for sound source localization, as well as signal processing and signal estimation concepts, such as prediction coefficients and residual signals. The latter types result in higher compression performance compared to binaural cues only, and, more importantly, they enable true 'transparency', which is impossible to achieve with binaural cues only.

But besides depending on the choice of parameters, the perceived quality of spatial audio coding techniques also depends on the parameter bit rate that was set at the encoder side. In principle, a large range of parameter bit rates can be applied, varying from a few kbps for extremely low bit-rate applications, up to 32 kbps or higher for near-transparancy.

MPEG Surround Standardization

The MPEG Surround standardization process was finalized in January 2007. MPEG Surround works with a large set of conventional 'core' coders such as MPEG-1 layer 2 and layer 3 (mp2 and mp3 in short), as well as AAC and HE-AAC. The parametric part of MPEG Surround, as well as the respective analysis and synthesis tools, re-use Parametric Stereo technology, extended with new algorithms specifically designed for multi-channel audio. The development of MPEG Surround is a joint effort of Philips, Coding Technologies, Agere and Fraunhofer.

MPEG Surround features

Here's a list of the most important features of MPEG Surround:

More information

(c) 2007 www.jeroenbreebaart.com