ISO/IEC 13818-3:1998 情報技術—動画および関連するオーディオ情報の一般的なコーディング—パート3：オーディオ | ページ 3

Introduction

ISO/IEC 13818 was prepared by SC29/WG11, also known as MPEG (Moving Pictures Expert Group). MPEG was formed in 1988 to establish a standard for the coded representation of moving pictures and associated audio stored on digital storage media.

ISO/IEC 13818 is published in three parts. 1 — systems — specifies the system coding layer of the standard. It defines a multiplexed structure for combining audio and video data and means of representing the timing information needed to replay synchronised sequences in real-time. 2 — video — specifies the coded representation of video data and the decoding process required to reconstruct pictures. 3 — audio — specifies the coded representation of audio data and the decoding process required to decode audio signals.

The technical changes in this 2nd edition compared to the first publication of ISO/IEC 13818-3 (1995) are:

• 1. In the first publication, certain combinations of dynamic crosstalk and prediction were not prohibited but not practically implementable. In this 2nd revision, these combinations are explicitly prohibited.
• 2. In the first publication, a low-pass filter was to be applied to the monophonic surround signal in matrix mode 2 (analogue surround mode). This filter is omitted in this edition, greatly simplifying the decoder.
• 3. The description of the syntax of the LFE channel was ambiguous. This description has been clarified.

Next to these technical changes, many editorial changes have been made, improving readability and clarity.

0.1 Extension of ISO/IEC 11172-3 Audio Coding to Lower Sampling Frequencies

In order to achieve better audio quality at very low bit rates (<64 kbit/s per audio channel), in particular if compared with ITU-T (formerly CCITT) Recommendation G.722 performance, three additional sampling frequencies are provided for ISO/IEC 11172-3 layers I, II and III. The additional sampling frequencies (Fs) are 16 kHz, 22,05 kHz and 24 kHz. This allows corresponding audio bandwidths of approximately 7,5 kHz, 10,3 kHz and 11,25 kHz. The syntax, semantics, and coding techniques of ISO/IEC 11172-3 are maintained except for a new definition of the sampling frequency field, the bitrate index field, and the bit allocation tables. These new definitions are valid if the ID bit in the ISO/IEC 11172-3 header equals zero. To obtain the best audio performance, the parameters of the psychoacoustic model used in the encoder have to be changed accordingly.

With these sampling frequencies, the duration of the audio frame corresponds to:

0.2 Low bitrate coding of multichannel audio

0.2.1 Universal multichannel audio system

A standard on low bit rate coding for mono or stereo audio signals was established by MPEG-1 Audio in ISO/IEC 11172-3. This standard is applicable for carrying of high quality digital audio signals associated with or without picture information on storage media or transmission channels with limited capacity.

The ISO/IEC 11172-3 audio coding standard can be used together with both MPEG-1 and MPEG-2 Video as long as only two-channel stereo is required. MPEG-2 Audio (ISO/IEC 13818-3) provides the extension up to 3/2 multichannel audio and an optional low frequency enhancement channel (LFE).

This part of ISO/IEC 13818 describes an audio subband coding system called ISO/MPEG-Audio Multichannel, which can be used to transfer high quality digital multichannel and/or multilingual audio information on storage media or transmission channels with limited capacity. One of the basic features is the backwards compatibility to ISO/IEC 11172-3 coded mono, stereo or dual channel audio programmes. It is designed for use in different applications as considered by the ISO/MPEG audio group and the specialist groups TG 10/1, 10/2 and 10/3 of the ITU-R (previously CCIR).

Multichannel audio systems provide enhanced stereo performance compared to conventional two channel audio systems. It is recognised that improved presentation performance is desirable not only for applications with

accompanying picture but also for audio-only applications. A universal and compatible multichannel audio system applicable to satellite or terrestrial television broadcasting, digital audio broadcasting (terrestrial and satellite), as well as other non-broadcasting media, e.g.,

seems to be very attractive to the manufacturer, producer and consumer.

0.2.2 Representation of multichannel audio

0.2.2.1 The 3/2-stereo plus LFE format

Regarding stereophonic presentation, specialist groups of ITU-R, SMPTE, and EBU recommend the use of an additional centre loudspeaker channel C and two surround loudspeaker channels LS and RS, augmenting the front left and right loudspeaker channels L and R. This reference audio format is referred to as"3/2-stereo" (3 front/ 2 surround loudspeaker channels) and requires the transmission of five appropriately formatted audio signals.

For audio accompanying picture applications (e.g. HDTV), the three front loudspeaker channels ensure sufficient directional stability and clarity of the picture related frontal images, according to the common practice in the cinema. The dominant benefit is the"stable centre", which is guaranteed at any location of the listener and important for most of the dialogue.

Additionally, for audio-only applications, the 3/2-stereo format has been found to be an improvement over two-channel stereophony. The addition of one pair of surround loudspeaker channels allows improved realism of auditory ambience.

A low frequency enhancement channel (in this part of ISO/IEC 13818 called LFE channel) can, optionally, be added to any of these configurations. The purpose of this channel is to enable listeners to extend the low frequency content of the reproduced programme in terms of both frequency and level. In this way it is the same as the LFE channel proposed by the film industry for their digital sound systems.

The LFE channel should not be used for the entire low frequency content of the multichannel sound presentation. The LFE channel is optional at the receiver, and thus should only carry low frequency sound effects, which may have a high level. The LFE channel is not included in any dematrixing operation in the decoder. The sampling frequency of the LFE channel corresponds to the sampling frequency of the main channels, divided by a factor of 96. This provides 12 LFE samples within one audio frame. The LFE channel is capable of handling signals in the range from 15 Hz to 120 Hz.

0.2.2.2 Compatibility

Extension from 2/0-stereo towards multichannel sound.

As a result of the widespread use of conventional two-channel stereo (2/0-stereo) reproduction, compatibility with existing 2/0-stereo sound reproduction systems or with existing matrixed surround sound receivers has to be maintained. This means that for many applications a basic stereo signal which contains an appropriate downmix of the audio information of the multichannel programme has to be transmitted together with the multichannel audio information. Appropriate downmix equations are given by equation pairs (1,2), (3,4), (5,6) and (7,8).

(1)

(2)

or

(3)

(4)

or

(5)

(6)

or

(7)

(8)

where jS is derived from LS and RS by calculation of the mono component. Then, a dynamic range compression and 90 degrees phase shift are applied to this component. The downmix (7,8) is suitable for existing matrixed surround decoders.

The format of an ISO/IEC 13818-3 bit stream is such that an ISO/IEC 11172-3 audio decoder properly decodes the basic stereo information according to one of the sets of downmix equations above (see 0.2.3.1). Compatibility with existing surround sound decoders by use of equations (7) and (8) has not been verified at the time of printing of this part of ISO/IEC 13818.

In the case of this part of ISO/IEC 13818, three different possibilities can be identified to provide to the user a basic stereo downmix together with the multichannel audio information:

• 1. Transmitting the 2/0-stereo sound inherently with the multichannel information in one bit stream in a backwards compatible way with ISO/IEC 11172-3, thus avoiding simulcast. This allows for the most efficient use of bit rate required for both, the 2/0-stereo and the multichannel audio signal. Additional advantages are that both programmes are strictly synchronized on a PCM audio sample basis, and that audio programme associated data carried in the ancillary data field of the MPEG-Audio bit stream have to be transmitted only once. The stereo downmix from the multichannel audio signal is handled by the ISO/IEC 13818-3 encoder. For this downmix, a number of matrix options according to equations (1) and (2) and equations (3) and (4) are provided by this part of ISO/IEC 13818 (see 2.5.2.13).
• 2. Simulcast of the multichannel audio signal, coded according to this part of ISO/IEC 13818, together with the 2/0-stereo signal coded according to ISO/IEC 11172-3. This solution requires two independent bit streams which can be multiplexed and transmitted by ISO/IEC 13818-1. The programme provider has to make provisions if a synchronization of both bit streams is required. Further, the simulcast option requires a significantly higher bit rate because instead of 5 channels in the case of 3/2 multichannel sound, altogether 7 audio channels have to be transmitted. However, the simulcast option allows for an individual, i.e. dynamic downmix to 2/0-stereo sound which can be controlled by a sound engineer.
• 3. Transmitting only the multichannel signal, by using the non-matrixed mode (downmix equation (5,6) ).

Each stereo decoder has then to be able to decode all the five channels, and to make a stereo downmix. Although the downmix can be applied before the filtering operation in the decoder, and the filter only needs to be done on two channels, this complicates the decoder significantly.

If compatibility with existing matrixed surround sound decoders is required, this part of ISO/IEC 13818 again provides three solutions:

• 1. To ensure a high efficiency regarding the bit rate required for both, the 3/2-multichannel and the matrixed surround signal, this surround signal can be transmitted in the backwards compatible stereo channel. The matrix-option '10' according to equations (7) and (8) provides an appropriate compatible signal which is transmitted in the basic stereo channels. A matrixed surround signal, suitable for existing matrixed surround decoders, can be obtained at the receiver by using an ISO/IEC 11172-3 two-channel decoder. The corresponding 3/2-channel output can be derived by using an ISO/IEC 13818-3 decoder.
• 2. A higher bit rate is necessary for simulcast of a matrixed surround signal using ISO/IEC 11172-3 and a 3/2-multichannel audio signal using this part of ISO/IEC 13818. This simulcast option allows for an independent mix of the matrixed surround signal which can be controlled by a sound engineer. The drawback of this solution is the additional bit rate necessary for transmitting 7 audio channels instead of only five channels if matrix-option '10' (see 2.5.2.13) is used.
• 3. Transmitting only the multichannel signal, by using the non-matrixed mode. Each stereo decoder has then to be able to decode all the five channels, and to make the downmix according to equation (7,8). Although the downmix can be applied before the filtering operation in the decoder, and the filter only needs to be done on two channels, this complicates the decoder significantly.

Downwards compatibility.

A hierarchy of audio formats providing a lower number of loudspeaker channels and reduced presentation performance (down to 2/0-stereo or even mono) and a corresponding set of downwards mixing equations are recommended in ITU-R Recommendation 775:"Multichannel stereophonic audio system with and without accompanying picture", November 1992. Alternative lower level audio formats which may be used in circumstances where economic or channel capacity constraints apply, are 3/1, 3/0, 2/2, 2/1, 2/0, and 1/0. Corresponding loudspeaker arrangements are 3/2, 3/1, 3/0, 2/2, 2/1, 2/0, and 1/0.

Backwards compatibility.

For several applications, the intention is to extend the existing 2/0-stereo sound system by transmitting additional audio channels (centre, surround) without making use of simulcast operation. This provision of backwards compatibility with existing receivers implies the use of compatibility matrices: the decoder of the previous generation must reproduce the two conventional basic stereo signals L'o/R'o, and the multichannel decoder produces the complete 3/2-stereo presentation L'/C'/R'/LS'/RS' from the basic stereo signal and the extension signals.

It is recognised that backward compatibility may not be required for all applications of MPEG-2 Audio.

Therefore, nonbackward compatible (NBC) audio coding systems free of the constraints of backwards compatibility are being evaluated for optional use with this part of ISO/IEC 13818.

0.2.2.3 Multilingual capability

Particularly for HDTV applications, multichannel stereo performance and bilingual programmes or multilingual commentaries are required. This part of ISO/IEC 13818 provides for alternative audio channel configurations in the five-channel sound system, for example a bilingual 2/0 stereo programme or one 2/0, 3/0 stereo sound plus accompanying services (e.g."clean dialogue" for the hard-of-hearing, commentary for the visually impaired, multilingual commentary etc.). An important configuration is the reproduction of commentary dialogue (e.g. via centre loudspeaker) together with the common music/effect stereo downmix (examples are documentation film, sport reports).

0.2.3 Basic Parameters of the Multichannel Audio Coding System

The transmission of the five audio signals of a 3/2 sound system requires five transmission channels (although, in the context of bitrate reduced signals, these are not necessarily independent). In order that two of the transmitted signals can provide a stereo service on their own, the source sound signals are generally combined in a linear matrix prior to encoding. These combined signals (and their transmission channels) are identified by the notation T0, T1,T2, T3 and T4.

0.2.3.1 Compatibility with ISO/IEC 11172-3

The ISO/MPEG-Audio Multichannel system provides full compatibility with ISO/IEC 11172-3. For a multichannel audio bit stream, backwards compatibility means, that an ISO/IEC 11172-3 audio decoder properly decodes the basic stereo information (see 0.2.2.2). Forwards compatibility means that an MPEG-2 multichannel audio decoder is able to decode properly an ISO/IEC 11172-3 audio bit stream.

The backwards compatibility is realised by coding the basic stereo information in conformance with ISO/IEC 11172-3 and exploiting the ancillary data field of the ISO/IEC 11172-3 audio frame (base frame, in the context of this part of ISO/IEC 13818) plus an optional extension frame for the multichannel extension.

The complete ISO/IEC 11172-3 audio frame incorporates four different types of information:

• Header information within the first 32 bits of the ISO/IEC 11172-3 audio frame.
• Cyclic Redundancy Check (CRC), consisting of 16 bits, just after the header information (optional).
• Audio data, for Layer II consisting of bit allocation (BAL), scalefactor select information (SCFSI), scalefactors (SCF), and the subband samples.
• Ancillary data. Due to the large number of different applications which will use this part of ISO/IEC 13818, the length and usage of this field are not specified.

The variable length of the ancillary data field enables packing the complete extension information of the channels T2/T3/T4 into the first part of the ancillary data field. If the MC encoder does not use all of the ancillary data field for the multichannel extension information, the remaining part of the field can be used for other ancillary data.

The bit rate required for the multichannel extension information may vary on a frame by frame basis, depending on the sound signals. The overall bit rate may be increased above that provided for in ISO/IEC 11172-3 by the use of an optional extension bit stream. The maximum bit rate, including the extension bit stream, is given by the following table:

This part of ISO/IEC 13818 describes the combinations of the basic Lo, Ro stereo of Layer I, II and III and the multichannel extension of Layer II mc and Layer III mc. The following combinations are possible:

0.2.3.2 Audio Input/Output Format

Sampling frequencies: 48, 44.1 or 32 kHz

Quantisation: up to 24 bits/sample PCM resolution

The following combinations of audio channels can be applied as inputs to the audio encoder:

• a) Five channels, using the 3/2 configurationL, C, R plus two surround channels LS, RS
• b) Four channels, using the 3/1 configurationL, C, R plus single surround channel S
• c) Three channels using the 3/0 configurationL, C, R without surround
• d) Five channels, using the 3/0 + 2/0 configurationL, C, R of first programme plus L2, R2 of second programme
• e) Four channels, using the 2/2 configurationL, R plus two surround channels LS, RS
• f) Three channels using the 2/1 configurationL, R with single surround channel S
• g) Two channels, using the 2/0 (or 1/0+1/0) configurationStereo (or dual channel mode) as in ISO/IEC 11172-3
• h) Four channels, using the 2/0 + 2/0 (or 1/0+1/0+ 2/0) configurationL, R (or channel I and channel II) of first programme plus L2, R2 of second programme
• i) One channel, using the 1/0 configurationSingle channel mode (as in ISO/IEC 11172-3)
• j) Three channels, using the 1/0 + 2/0 configurationSingle channel mode (as in ISO/IEC 11172-3) plus L2, R2 of second programme

The different combinations of audio input signals are encoded and transmitted within the up to five available transmission channels T0, T1, T2, T3 and T4, of which channels T0 and T1 are the two basic channels of ISO/IEC 11172-3 and convey the backwards compatible signals Lo and Ro. Transmission channels T2, T3 and T4 together form the multichannel extension information, which is compatibly transmitted within the ISO/IEC 11172-3 ancillary data field and an optional extension bit stream.

After multichannel decoding, the up to five audio channels are recovered and can then be presented in any convenient format at the choice of the listeners:

• a) Five channels, using the 3/2 configurationFront: Left (L) and right (R) channel plus centre channel (C)Surround: Left surround (LS) and right surround (RS)
• b) Four channels, using the 3/1 configurationFront: Left (L) and right (R) channel plus centre channel (C)Surround: Mono surround (S)
• c) Three channels using the 3/0 configurationFront: Left (L) and right (R) channel plus centre channel (C)Surround: No surround
• d) Four channels, using the 2/2 configurationFront: Left (L) and right (R) channelSurround: Left surround (LS) and right surround (RS)
• e) Three channels, using the 2/1 configurationFront: Left (L) and right (R) channelSurround: Mono surround (S)
• f) Two channels, using the 2/0 configurationFront: Left (L) and right channel (R)Surround: No surround
• g) One channel output, using the 1/0 configurationFront: Mono channel (Mo)Surround: No surround

A low frequency enhancement channel can, optionally, be added to any of these configurations, except for the 1/0 configuration.

Outputs may be required to provide discrete signals, or may be combined in accordance with downward mixing, or upwards conversion equations, as defined in ITU-R Recommendation 775.

0.2.3.3 Composite Coding Modes

Dynamic Transmission Channel Switching

In order to provide a better orthogonality between the two compatible signals T0 and T1, and the three additionally transmitted signals T2, T3 and T4, it is necessary to have flexibility in the choice of the channels T2, T3 and T4. This part of ISO/IEC 13818 allows, independently for a number of frequency regions, the selection of a number of combinations of three out of the five signals L, C, R, LS, RS to be transmitted in T2, T3 and T4.

Dynamic Crosstalk

According to a binaural hearing model, it is possible to determine the portion of the stereophonic signal which is irrelevant with respect to the spatial perception of the stereophonic presentation. The stereo-irrelevant signal components are not masked, but they do not contribute to the localisation of sound sources. They are ignored in

the binaural processor of the human auditory system. Therefore, stereo-irrelevant components of any stereo signal (L, C, R, LS or RS) may be reproduced via any loudspeaker, or via several loudspeakers of the arrangement, without affecting the stereophonic impression. This can be done independently for a number of frequency regions.

Adaptive Multichannel Prediction

In order to make use of the statistical inter-channel dependencies, adaptive multichannel prediction is used for redundancy reduction. Instead of transmitting the actual signals in the transmission channels T2, T3, T4, the corresponding prediction error signals are transmitted. A predictor of up to 2nd order with delay compensation is used.

Phantom Coding of Centre

Due to the fact that the human auditory system uses only intensity cues of the audio signal for localisation at higher frequencies, it is possible to transmit the high frequency part of the centre channel in the front left and right channels, constituting a phantom source at the location of the centre loudspeaker.

0.2.3.4 Encoder and Decoder Parameters

Annex E

Ancillary Data Use

Introduction

A number of existing applications of MPEG Audio, including international standards (e.g. DAB [1], ITU-T J.52 [2]) have defined formats for the ancillary data field, according to their specific requirements. In this Annex, some examples are illustrated, that might be of interest for future applications.

Each ISO/IEC 13818-3 frame may contain a number of ancillary data bytes. This data can be carried in two separate fields of an ISO/IEC 13818-3 encoded frame. One field is located at the end of the base frame, in order to be compatible with the ancillary data definition of ISO/IEC 11172-3, the other is located at the end of the extension frame.

The most common use of ancillary data is Programme Associated Data (PAD), data intimately related to the audio signal.

Typical Programme Associated Data

Typical examples of Programme Associated Data are the indication of music or speech (Music/Speech flags), programme related text (ITTS [1]), Universal Product Code/European Article Number (UPC/EAN [1]]), special commands to a receiver/decoder which are provided synchronously to the audio programme, and Dynamic Range Control information (DRC). The DRC signal can be optionally used in the receiver to compress the dynamic range of the audio signal. This is an example of such data which would become useless if delayed when queuing in a shared data service.

All functions provided by PAD and the length of the PAD fields are user definable. Therefore, it is not mandatory to send any information in the PAD field.

Dynamic Range Control

It has long been recognised that for many listeners in less than perfect environments it is impractical to make use of the full dynamic range which may be carried by a digital audio signal. Methods of carrying data in a coded bitstream in order to allow a limitation of the reproduced dynamic range of an audio signal have already been defined for ISO/IEC 11172-3 Layer II for Digital Audio Broadcasting (DAB [1]).

With the help of the Dynamic Range Control (DRC) the receiver may reduce the dynamic range of the audio signal. The purpose of this is to adapt the dynamic range of the audio signal to listening either in a noisy environment or to an audio source which has a too high dynamic range (typical movie sound tracks) for domestic listening. ISO/IEC 13818-3 decoders may, optionally, provide such a compression of the audio dynamic range, using a process which either derives its control information from the audio signal itself or from a suitable DRC signal transmitted in the ancillary data field. It is an option for the programme provider to transmit a DRC signal; it is not a requirement of the system.

In the DAB specification, part of the extra data carried with the audio (the"F-PAD") can, amongst other things, carry a six bit DRC data field which is to be used to vary the gain applied to the reproduced audio signal. In current proposals [1], when Dynamic Range Control is signalled the six bits represent a gain to be applied to the recovered audio in the range of 0 — 15.75 dB in steps of 0.25 dB. The step size of 0.25 dB has been found by experimentation to be largest acceptable to provide smooth gain control during slow gain changes in classical music. The upper limit of 15.75 dB as the maximum signalled gain is thought to be adequate to allow suitable dynamic range reduction under not too difficult listening conditions. If a further reduction in dynamic range is required due to extremely adverse conditions, the signalled values can be scaled and the increase in step size is unlikely to be audible. If Dynamic Range Control data is being sent, the six bit value is required to be sent once every 24 ms. This represents a bit usage of 250 bits/sec (not counting the overhead required to signal the use of the DRC data).

Music/Speech Indication

These two flags indicate whether the transmitted sound consists of music or speech. The receiver may use this information to control any sound processing circuitry. One special combination of the flags signals that no indication is given. The Music/Speech indication typically requires two bit, repeated about 10 times per second.

Commands to a Receiver/Decoder

A channel can be provided to convey, synchronously to the audio signal, special commands to the receiver/decoder. Such commands may be used, for instance, to trigger the read out of a picture from a buffer memory that was filled, asynchronously, in advance. This channel is able to carry a few bytes within 0.2-0.5 second, at irregular intervals.

Programme related text

To elucidate the transmitted audio signal — a song, a programme item — coded text may be carried together with the audio. This text may be made on-site by the programme provider, it may be read from digital pre-recorded software and relayed more or less transparently, or various sources can be combined. The channel capacity required for text is dependent on how comprehensive and attractive the service is made.

In-house Information

Channels can be provided for both short, synchronous commands and for long strings of asynchronous data. The meaning of these commands is intended for internal use within a specific application only.

References: