Organisation internationale de normalisation

Continued discussion of CE on Tontal Components

Lukasz Januszkiewicz, Zylia, gave a presentation on the CE on Tonal Components

The goal of this CE setup was to show that the CE technology did not make “typical” signals worse, and it does not influences the spatial audio quality. The listening test results for 9.0 item showed that it can still improve the quality for multichannel signals. The pooled data differential results for stereo test were:

1. 
   2 items better, mean better, 2 worse.
   

The presenter noted that one of the worse item was phi7, which is pitch pipe, and felt that there was a failure in the decoder synthesis algorithm. 

The presenter proposed a workplan for a next step.

At the 111^th MPEG meeting Zylia presented CE performance information of the following:

There were two system configuration tested, 16 kb/s and 20 kb/s, both for mono test items provided by Zylia. The listening test results were as follows:

Absolute scores:

• 
  16 kb/s – 1 item better and the mean score better,
  
• 
  20 kb/s – 1 item better and the mean score better.
  

Differential scores:

1. 
   16 kb/s – 8 items better and the mean score better,
   
2. 
   20 kb/s – 9 items better and the mean score better.
   

At the 112^th meeting Zylia presented CE performance information of the following:

• 
  Bitrate 0.6 kb/s/channel
  
• 
  Complexity 0.04 WMPOPS per channel, which is less than 1% with respect to total 3D Audio decoder.
  

The goal of this CE setup was to that the CE technology did not make “typical” signals worse, and perhaps showed improvement.

Differential

• 
  2 items better, one worse
  

The presenter noted that the one item worse was phi7, which is pitch pipe, and felt that there was a failure in the decoder synthesis algorithm.

The presenter proposed a workplan for a next step.

Issues to be addressed:

• 
  Does delay need to be reduced? Experts state more than 0.5 seconds encoder total delay would be an issue, so the CE's encoding delay of 10240 samples is accepted.
  
• 
  System configuration for next phase of listening test was stated as follows:
  
  • 
    20 kb/s stereo
    
  • 
    items submitted by Zylia (including phi7 and Speech_Over_Music_4)
    
  • 
    for reference the USAC RQ encoder will be used, delivered by FhG in the previous listening tests
    
• 
  Huffman codes – experts stated that additional Huffman code tables that can increase efficiency of the CE will not be an issue.
  
• 
  Other issues concerning the sinusoidal synthesis complexity and bitstream syntax were clarified during further discussions with experts:
  
  • 
    control information for the operation of the tool can be transmitted in an mpegh3daExtElement container
    
  • 
    alternate methods for synthesizing the high frequency sinusoidal components will be further investigated
    
  • 
    complexity due to additional QMF analysis filterbanks may be controlled by limiting the number channels on which the HFSC is applied
    
  • 
    delay due to additional QMF analysis filterbanks will not add to the overall decoder delay
    

Adrian Murtaza, FhG-IIS, presented

m36544

Proposed Updates to MHAS

Adrian Murtaza, Herbert Thoma, Harald Fuchs, Achim Kuntz, Max Neuendorf, Andreas Niedermeier, Stephan Schreiner

The contribution proposes new functionality in MHAS:

• 
  Global CRC check
  
• 
  Audio sample truncation
  

The audio sample truncation capability permits 3D Audio decoder output to have sample-accurate alignment with video frame rate, and permits simple audio/visual program splicing.

There was some discussion on where the truncation occurs: after the mixing stage or at the end of the decoding process.

Subject to the clarification of where in the decoder the truncation can occur, it was the consensus of the Audio subgroup to incorporate this technology into MPEG-H Phase 2.
Michael Kratschmer, FhG-IIS, presented

m36586

Metadata Updates to MPEG-H 3D audio

Simone Fueg, Jan Plogsties, Michael Kratschmer

The contribution notes that ITU-R has standardized, as communicated to MPEG in its liaison statement to this meeting, metadata for audio broadcast:

• 
  Broadcast WAV 64bit (BW64)
  
• 
  Audio Definition Model (ADM)
  

The Chair noted that ADM work was initiated in EBU, which presented this work to the Audio subgroup at a previous MPEG meeting.

The primary motivation of the contribution is to permit 3D Audio to ingest the ITU-R metadata without loss of information.

The contribution proposes support for the following metadata changes or additions: