Organisation internationale de normalisation

Discussion

There are two alternatives:

1. 
   Remove 960 block length from the AAC and HE-AAC profile.
   
2. 
   Keep the current profiles and request that manufacturers implement both block lengths. This education effort must include enhancing conformance and reference software to support 960 block length.
   

Dave Singer, Apple, noted that the real goal is to minimize the confusion in the marketplace when DAB 960 block length content becomes available in the “Internet download” marketplace. Experts should investigate whether DAB bitstreams can be saved to an external storage device.

After considerable discussion, all agreed that one way forward is:

• 
  Minimize the number of 960 block length bitstreams
  
• 
  Maximize the number of new dual-block block length decoders
  

And that another way forward is:

• 
  Restrict decoders to 1024 block lengths (960 for DAB and similar special applications)
  
• 
  Educate 960 block length content producers that they must transocode to 1024 block length in order to make content available outside their integrated systems.
  

Yasushige Nakayama, NHK, presented

m15598

Proposal of work plan for a 3D multichannel audio configuration

Yasushige Nakayama

This contribution notes external standards that specify mapping from audio channels to loudspeakers. It further presented relevant points from the draft technical broadcasting standard from ARIB (Japan), which include 22.2 channel broadcast using AAC or ALS coding technology. To support this initiative, NHK requests that MPEG extend AAC-LC to signal a mapping of up to 22.2 audio channels to the associated loudspeakers. In order for this to align with the ARIB schedule, they request FDIS by October 2009.

Audio Chair noted that it would be most desirable to have a general audio channel to loudspeaker mapping, and that a BIFS scene might support such a channel mapping.

Karlheinz Brandenburg, Fraunhofer Ilmenau, noted that there are many loudspeaker configuration that should be addressed, including:

• 
  5.1
  
• 
  7.2
  
• 
  16 channel (US theatres)
  
• 
  22.2 (NHK)
  
• 
  Higher-order Ambisonics
  
• 
  Wavefield synthesis
  

Juergen Schmidt, Thomson, made a presentation on how Audio BIFS might be used to address audio channel to output loudspeaker mappings. BIFS is a tree-based organization. The BIFS stream specifies the node tree and the mapping and further manages the connections from compressed stream to audio composition.

A rather free-form set of notes from that presentation follows:

AudioChannnelConfig node

Labels input channels

See Channel Configuration Table in 14496-3

Parametric channel oriented configuration

Generic format with no limitations

BIFS scene decoder can be unbounded in complexity. However profiles and levels can be used to limit complexity to be very low. This could use the following

• 
  Sound2D
  
• 
  AudioChannelConfig
  
• 
  AudioSource
  

Decoder instances

Very simple BIFS decoder

See Abv3_Aac06a.txt BIFS conformance data

Way forward

Define new profile and level to meet this application and keep bitrate and complexity low.

Streaming format

• 
  Use MPEG-4 Systems
  
• 
  Use “Carousell” to enable break-in
  

The Audio Chair noted that it would be appropriate to ask the Systems Chair if Laser can support 22.2 channel audio.

It was decided to create an AhG mandate to continue this work, consisting of at least the following interested experts:

Jurgen Schmidt, Thomson

Schuyler Quackenbush, Audio Research Labs

Karlheinz Brandenburg, Fraunhofer Ilmenau,

Yasushige Nakayama, NHK

Tilman Liebchen, LG