INTERNATIONAL ORGANISATION FOR STANDARDISATION
ORGANISATION INTERNATIONALE DE NORMALISATION
ISO/IEC JTC1/SC29/WG11
CODING OF MOVING PICTURES AND AUDIO
ISO/IEC JTC1/SC29/WG11
N3794
January 2001, Pisa
Source: Audio Sub group
Title: Call for Proposals for New Tools for Audio Coding
| This web page contains only excerpts of the document. The complete document is available as PDF file. |
In mid-1999 International Standard ISO/IEC 14496-3, MPEG-4 Audio Version 1 issued and in early 2000 ISO/IEC 14496-3 / AMD1, MPEG-4 Audio Version2 issued. Numerous tests have been conducted by MPEG (see references) to verify that the MPEG-4 standard contains state of the art technology. However, WG11 is always interested in new developments which may provide improvements over the existing MPEG-4 standard and which may lead to extensions of MPEG-4 or to new work items.
For this reason, at the 53rd MPEG meeting, in Beijing, MPEG issued a Call for Evidence Justifying the Testing of Audio Coding Technology (N3641). Evidence submitted in response to the Call was examined at the 55th MPEG meeting, and it was determined that there was technology that might improve upon the MPEG-4 standard.
Therefore WG11 issues with this document a call for proposals of new audio coding technology.
WG11 is interested in technology that
All proposals of technology have to fulfil the requirements that are defined in ANNEX I, which is to supply a technical description and evidence of the performance of the proposed technology.
The following steps are planned for the standardisation of the new technology:
Conditions for the specific technologies in the call follow.
Bandwidth extension is a tool that is not yet available within MPEG-4. This call asks for technology that addresses bandwidth extension of one or both of:
A single technology that addresses both of these signals is preferred. This technology shall be both forward and backward compatible with existing MPEG-4 technology. In other words, an MPEG-4 decoder can decode an enhanced stream and a new technology decoder can decode an MPEG-4 stream. There are two possible configurations for the enhanced stream: MPEG-4 AAC streams can carry the enhancement information in the DataStreamElement, while all MPEG-4 decoders can accept a MPEG-4 Elementary Stream and a second Elementary Stream containing the enhancement information.
In the formal verification test it is determined whether the developed technology provides a significant increase in coding efficiency in comparison to MPEG-4. The developed technology shall satisfy both of the following two criteria. (The MUSHRA test methodology will be used).
The target bit-rate for the proposed coder is approximately 24 kbit/s per channel for general audio, and the target bit-rate is approximately 8 kbit/s per channel for speech.
As a result of optimising, the quality of the compatible part might be less than that of a regular encoder. The quality of the core coder shall be compared to MPEG-4 operating at a bit-rate 25% lower than the target bit-rate.
The MPEG-4 standard already provides a parametric coding scheme for coding of general audio signals for low bit-rates (HILN). This call asks for technology that addresses parametric coding of general audio signals for the higher quality range, to extend the capabilities currently provided by HILN. Whenever possible this technology should build upon the existing MPEG-4 HILN or other MPEG-4 technology.
In the formal verification test it is determined whether the developed technology provides a significant increase in coding efficiency in comparison to MPEG-4. The developed technology shall satisfy both of the following two criteria. (The MUSHRA test methodology will be used).
The target bit-rate for the proposed coder is approximately 24 kbit/s per channel.
Register by 15 May, 2001 an intention to participate in the Call. Register by sending an email to Schuyler Quackenbush (Chairman of the MPEG Audio Subgroup, srq@research.att.com). Email should indicate contact names, company and the technology that will be proposed (e.g. bandwidth extension for general audio, bandwidth extension for speech, or parametric coding for general audio).
Submit by 1st June, 2001, the following: the bitstreams, decoders and decoded sound files (*.wav) associated with the proposed algorithms. Decoders shall be delivered as executables on x86 Linux or Win32 platforms. Proponents that have already submitted these materials in response to the Audio Call for Evidence (N3641) do not need to submit them again.
Submit as contributions to the July MPEG meeting:
All proponents need to submit a description. Proponents that have already submitted evidence in response to the Audio Call for Evidence (N3641) are not required to submit this again. Proponents that are MPEG members shall register these documents as contributions to the July MPEG meeting and send title and author information to Schuyler Quackenbush prior to the time of the close of the contribution registry. Proponents that are not MPEG members shall email the documentation to Schuyler Quackenbush prior to 27 June, 2001, so that he can register them as contributions. The proposer's documents should be written in Microsoft Word.
Attend the July MPEG meeting (details on meeting location and date will be communicated via email). It is strongly urged that experts familiar with the proposed technology attend in order to allow discussions on details of the proposals.
At the July MPEG meeting timetables and procedures will be defined for conducting three tests that will compare the proposed technology, one test for each of bandwidth extension of general audio signals (at approximately 24 kb/s), bandwidth extension of speech signals (at approximately 8 kb/s, if sufficient evidence is provided) and parametric coding of general audio signals (at approximately 24 kb/s). The MUSHRA test methodology will be used and the timetable for submissions of proponent materials for the test will be determined at the July MPEG meeting.
The best technology, as identified by the comparative tests, will be Verification Model 1 and be the basis for subsequent core experiments. Proponents whose technology is selected as Verification Model 1 and all proponents participating in the core experiment process shall supply a detailed description of their technology. Core experiments will be conducted according to Core Experiment Methodology for MPEG-4 Audio, N1748.
In core experiments, the performance of equivalent MPEG-4 blocks will be checked. In the case of comparable performance, the existing MPEG-4 technology will have preference.
The performance of the new technology will be measured via a formal subjective test, to be carried out prior to the Committee Draft stage of the standardization process. The acceptance criteria (enumerated above) must be met in order for the technology to progress in the standardization process.
Heiko Purnhagen 07-Feb-2001