Organisation internationale de normalisation

26Task group activities

26.1Joint Meeting

26.1.1With Requirements on Audio for HEVC (Wed 1400-1500)

Akio Ando, NHK, presented

m19381

Conversion of multichannel sound signal for introduction of 3D audio into home

Akio Ando

The contribution reviewed NHK’s 22.2-channel presentation format. NNK plans to implement a 22.2 channel broadcast service in 2020, however, it is clear that not every home can have or would choose to have a 22.2 channel listening setup.

One means to make the 22.2 broadcast service compatible to possible home theatre setups is to downmix the 22.2 signal to a lower number of channels (e.g. 8.1). The proposed conversion method creates a phantom source of the original sound signal at the corresponding downmix loudspeaker position. It further partitions the problem into 3x3 sub-matrices, such that the phantom source is computed from only the nearest physical loudspeakers.

A critical issue in the downmix problem is to determine the minimum number of speakers required to obtain a “similar” experience to the 22.2 channel presentation.

NHK used the BS.1116 test methodology to assess the relative quality of the following presentation systems:

REF 22 channel

Sys1 10 channel

Sys2 8 channel

Sys3 6 channel

For each number of downmix channels (10, 8, 6), three different speaker layouts were tested. LFE channels were neglected. Test items were 10 s in duration and more than 30 subjects participated in the test. Based on the test results, it was concluded that 8 channels was the minimum that received a score comparable (4.5 out of 5) to the 22.2 channel presentation.

An additional subjective test assessed the size of the “sweet spot” of the audio presentations provided by the various number of loudspeakers.

The group discussed what would be the opportunities for standardization for the scenario presented by NHK:

• 
  Efficient representation of 22.2 channel audio program.
  
• 
  Efficient downmix from 22.2 to the target number of speakers in the home setup. While this is a post-processing step, it may involve interoperability if downmix guidance information is transmitted.
  

The Convenor suggests that the appropriate way forward is to, first, develop requirements that address at least the NHK problem that the contribution describes. The requirements should be reviewed and refined, and then Audio experts can assess the extent that current MPEG technology may address those requirements.

Werner Oomen, Philips, noted that it would be possible to transmit an 8-channel program and then upmix that to 22-channel program. The presenter remarked that he has investigated this and concluded that the upmix scenario cannot deliver a sufficiently good audio experience.

Convenor noted that the problem might be stated as: map microphone number and position at source to loudspeaker number and positions at receiver. Audio Chair noted that objects or source are available if there is an actual recording with microphones, but is lost in the course of post-production in e.g. a typical movie soundtrack.

It was decided that the next step would be for Audio experts will draft a problem statement as an output from this meeting.