International organisation for standardisation

With Requirements on Audio for HEVC (Thu 1000-1100)

Since the author of this contribution could not attend the 96^th meeting, the Audio Chair presented

m19570

22.2 multichannel sound material for testing 3D coder

Akio Ando

The contribution briefly described the setup for several 22.2. channel records make by NHK. The objective was to record sound material with (1) a clear sound direction, (2) rich sound envelopment, (3) movement of sound with power, and (4) a pure and clear timbre. To obtain such a sound material, NHK selected four sound sources: (1) a church bell, (2) a pipe organ in a very reverberant room, (3) a steam locomotive, and (4) a Japanese Edo wind bell.

NHK is willing to make these recordings available to companies working on the “Audio of HEVC” exploration and any standard that may arise from that.

Jeongil Seo, ETRI, presented

m20107

Comments on High Efficiency Video Coding

Jeongil Seo, Kyeongok Kang

The contribution reviewed the concept of sound field reconstruction (SFR) and noted various technologies that can be employed for SFR

• 
  Vector Based Panning
  
• 
  Wave Field Synthesis
  
• 
  Higher Order Ambisonics
  
• 
  Least Squared Methods
  

It noted that there is ongoing work in ITU-R SG6 WP6C on “Multichannel Sound Technologies in Home and Broadcasting”

There was good discussion on SFR and the opportunities for standardization. The Requirements Chair noted that any potential requirements should be on a transmission format and on presentation format

Juergen Herre, FhG, noted that there are three representation schemes discussed in the contribution:

• 
  Channel based (e.g. NHK 22.2)
  
• 
  Object based (e.g. objects in sound scene)
  
• 
  Scene based (e.g. Ambisonics)
  

It was noted that ambisonics is very flexible with respect to scene rendering and loudspeaker placement, but has a very restrictive “sweet spot.” Higher-order ambisonics can extend the “sweet spot”of the perceived sound scene.

Audio Chair noted that there is a trade-off between channel-based and object-based paradigms:

• 
  Channel-based
  
  • 
    potentially simple and straightforward recording and encoding/decoding
    
  • 
    can guarantees the intended user experience if and only if the user has the correct loudspeaker placement
    
• 
  Object-based
  
  • 
    requires a potentially more complicated recording processing
    
  • 
    encoding/decoding (of object signals) is straightforward
    
  • 
    reconstructing the multi-channel user experience in the decoder is potentially very complex
    
  • 
    can potentially support arbitrary speaker placement in the user’s setup
    

Juergen Herre, FhG, noted that WFS is typically a 2D technology, i.e. a ring of loudspeakers in the horizontal field, and what we desire is 3D Audio, which may encompass height and also distance. He further noted that it would be very desirable if the presentation could be “best-effort” in that it adapts as best as possible to the users loudspeakers setup.

As a conclusion of the joint meeting, it was decided to add use cases and possible requirements appropriate to the object based representation to the “Audio for HEVC” document and, in addition, to add an section discussing