International organisation for standardisation organisation internationale de normalisation

4.2.2MPEG-D Spatial Audio Object Coding

It was confirmed that a response to the German NB could be to incorporate the requested technology into SAOC over the course of the next two MPEG meetings and to re-issue the SAOC FCD shortly after the 89^th MPEG meeting.

4.2.3MPEG-D Unified Speech and Audio

Markus Multrus, FhG, presented

m16321

Proposed Additions to and Corrections of the USAC Reference Software

Stefan Bayer Markus Multrus

This contribution proposes to add a software implementation of the Time Warped MDCT module for the encoder.

As a second point, it identifies a bug in

• 
  the buffer reset/update flag
  
• 
  TW-MDCT incorrect delay handling
  

These bugs have no impact on the RM0 bitstreams since RM0 never used these modes.

Finally, it notes that there are currently two MDCT implementations in the RM0 code base. It proposes to merge these two such that only one code base is used.

It was the consensus of the Audio Subgroup to accept these corrections and additions into the RM reference software.

The Chair presented

m16434

Draft Revised Audio CE Methodology

Schuyler Quackenbush

The Chair highlighted areas of the document that need study and input from the group. It was agreed in principal that it is of paramount importance that the revised document serve USAC, but that it also be as generic as possible while at the same time carry forward the minimum of “special case conditions” for old work items (e.g. lossless coding).

Kristofer Kjörling, Dolby, presented

m16314

Progress report on harmonic transposer CE for the USAC work item

Kristofer Kjörling Max Neuendorf

The contribution reports on status of this CE. The proponents expect to have a complete CE submitted to the next MPEG meeting.

Kei Kikuiri, NTT DoCoMo, presented

m16397

Core Experiment Proposal on the eSBR module of USAC

Kei Kikuiri Kousuke Tsujino Nobuhiko Naka

This contribution describes a CE to add a new tool to the eSBR module. It notes that, currently, the eSBR module can only adjust the temporal enveolope in the SBR sample domain with a granularity of 2 subband samples. This is not as fine a granularity as is provided in the LP time-domain coding tool and may be an issue when coding speech signals.

What is proposed is enhanced Temporal Envelope Shape, similar to what is present in MPEG Surround. A listening test shows the performance of the new tool. It showed improvement for 1 of 4 speech items in the test at the 95% level of significance. It reports that the TES tool requires an additional 2 bits per SBR envelope per channel (approximately 90 bps), and has some modest increase in complexity. The control of TES was done as a stand-alone module that operated on the RM0 bitstreams.

The Chair asked the group which operating modes and test items should be used to assess the performance of this CE. The Chair asked for which test items was Beta different from zero (i.e. the TES tool was active).

Hyunkook Lee, LG, presented

m16446

Core experiment proposal on arithmetic coding

Sungyong Yoon Hyunkook Lee Younghee Choi

This contribution proposes a CE in which the USAC global gain and differential scale factors are coded using arithmetic entropy coding rather than Huffman entropy coding. It reported an average bitrate savings of 0.39 percent.

The Chair asked what fraction of the reported gain is due to compressing the 8-bit PCM global gain. A question was raised as to the corpus that was used to train the arithmetic coder. The Chair suggested that, in the workplan, that the CE proponent supplies a corpus of material to the RM proponent to encode and then supply the resulting bitstreams to the CE proponent.

Kristofer Kjörling, Dolby, noted that there is potentially room for major improvements in the signal processing in USAC.

The proposal will be discussed again during the week when additional information is available (i.e. 1) exclude global gain and 2) reset between coding scale factors and spectral coefficient.