Organisation internationale de normalisation

26.2Task Group discussions

26.2.1MPEG-2, MPEG-4, MPEG-7, Audio Conformance, Reference Software, MPEG Surround

Takehiro Moriya, NTT, presented

m19197

Proposed draft corrigendum of ISO/IEC 14496-3:2009 for ALS floating point data corrections

Noboru Harada, Yutaka Kamamoto, Takehiro Moriya, Tilman Liebchen, Yuriy Reznik,

The contribution is the text for a DCOR against ALS that is proposed to issue at this meeting. It incorporates text from previous Defect Reports and adds some additional corrections and clarifications to the floating-point mode.

It was the consensus of the Audio subgroup to issue the contribution text as ISO/IEC 14496-3:2009/AMD 2:2010/DCOR 1:2010, Floating point data corrections.

Ferenc Kraemer, Dolby, presented

m19259

Additional information on MPEG Surround conformance testing

Andreas Hölzer, Christian Ertel, Markus Lohwasser, Michael Härtl, Ferenc Kraemer, Frans de Bont

The contribution reviewed an issue that was raised at the 94^th MPEG meeting: that various decoders (high quality, low power, floating point, fixed point) produce decoded waveforms that differ by as much as 200 lsb. The Audio subgroup looks forward to more information on this issue, perhaps at the 96^th meeting.

Heiko Purnhagen, Dolby, presented

m19286

Information on MPEG Surround Reference Software

Ferenc Kraemer, Leif Sehlstrom, Heiko Purnhagen, Christian Ertel, Andreas Hölzer, Johannes Hilpert

The contribution gives information that will be useful in preparing the DoC on N11501, ISO/IEC 23003-1:2007/AMD 2:2008/Cor 2.2.

Specific issues addressed are:

• 
  Support for 7-2-5 coding mode is missing from the reference software
  
• 
  The temporal shaping tools STP/GES are incorrectly enabled for the 5-1-2 mode.
  
• 
  Numerous editorial corrections were identified (e.g. consistent file names and help messages on various platforms)
  

The presenter felt that a significant editing period is required to finalize corrections in the reference software, and with the 96^th meeting approximately 2 months from now, it was agreed that the DoC be processed and the Cor issued at the 96^th meeting.

Leon Terentiv, FhG, presented

m19288

Corrections of the parameter processor for MPEG SAOC

Jonas Engdegård, Heiko Purnhagen, Oliver Hellmuth, Jürgen Herre, Cornelia Falch, Leon Terentiv, Johannes Hilpert, Andreas Hölzer, Werner Oomen

The contribution proposes a resolution to the issue raised at the 94^th meeting – that SAOC Karaoke mode could be significantly improved, but that all details of the parameter control were not yet known.

Two problems have been addressed:

• 
  OLD renormalization – the current specification does not re-normalize the remaining OLDs when one of the sound objects is “eliminated,” as with Karaoke mode. In the case that none of the remaining OLDs is equal to 1.0, this violates energy preservation. The contribution proposed semantics to correct this.
  
• 
  Eigenvalue limiting value – the current specification uses a threshold of 1/30. Based on further experimentation, particularly of both re-mix and karaoke use cases, the contribution proposes to change the threshold to 1/80.
  

Results of two listening tests were presented. In the first test addresses the Karaoke application and shows that for some items, the quality is already good and the proposed changes provide no change in quality. However, for other items, the OLD re-normalization provides a 20 MUSHRA point increase in performance, and both re-norm and eigenvalue limiting provides 35 MUSHRA points increase in performance, both of which are significant at the 95% level of significance. In the second test, which addresses the re-mix application, the proposed change caused no change in subjective performance, at the 95% level of significance.

Mohammed Raad, RaadTech, noted that it would be beneficial to advise industry that a DCOR against SAOC will issue at the 96^th meeting (such that they might hold off on finalizing any SAOC implementations). The Chair will draft a resolution to bring this message to national bodies and their member companies.

It was the consensus of the Audio subgroup to incorporate the contribution text into the Defect Report from the 94^th meeting and issue this as a Defect Report from this meeting.

The FhG test results may be verified by LG by the next meeting at which time a reissue of the SAOC verification test report might be considered.