International organisation for standardisation

MPEG-D Unified Speech and Audio Coding

Maintenance

Max Neuendorf, FhG, presented

m20034

Corrections to Reference Software and DIS of USAC

Max Neuendorf

The contribution divides the proposed corrections into three groups:

• 
  Corrections to the DIS text
  
• 
  Corrections to the Reference Software
  
• 
  Corrections to both DIS text and Reference Software
  

Corrections to the DIS text

• 
  TS Decorrelator
  
• 
  Syntax – add table that indicate that certain syntactic elements are in a clause of the MPEG Surround standard.
  
• 
  Add figure for MPS212 and SBR showing “normal” operation
  
• 
  FAC – change fac_length from fixed number to e.g. coreCoderFrameLength/8
  
• 
  Remove MPEG-4 AudioSpecificConfig() changes. They are moved to a possible MPEG-4 amendment
  
• 
  Correct alignment of LP mode adaptive codebook filter
  
• 
  Other small editorial corrections
  

Corrections to the Reference Software

• 
  Coarse ICC parameter dequantization is not aligned with text
  
• 
  Adopt corrections to MPEG Surround
  
• 
  Correct SBR buffer allocation size in Inter-TES
  
• 
  Correct handling of DC/Nyquist values in FFT used in SBR
  
• 
  Cross-products only for pitch>0 (i.e. never for unvoiced frames)
  
• 
  Buffer initialization in QMF transposer
  

Corrections to both DIS text and Reference Software

• 
  Return to state in which bsFixedGainDMX is available. It was discovered that this is needed for current USAC operation
  
• 
  Correct bsXXXDataMode “default” and “interpolation” modes. This aligns USAC MPS212 with fixes being made to MPEG Surround reference software.
  
• 
  Harmonize TNS_MAX_LENGTH:
  
  • 
    4-bit word length
    
  • 
    max of order 15 filter (i.e. using entire space of what can be signalled)
    
• 
  Coding of first FD scale factor
  
  • 
    shall be zero and shall not be signalled (since it is equal to GlobalGain)
    
• 
  Include TCX FAC into regions to which the bass post-filter is applied.
  

It was the consensus of the Audio Subgroup to adopt into the Study on DIS all changes proposed in this contribution.
Roch Lefebvre, VoiceAge, presented

m20031

Additions to USAC specification

Roch Lefebvre, Redwan Salami,Philippe Gournay, Bruno Bessette

The contribution proposes text to add to USAC specification. Some of the proposed text permits the removal of references to the 3GPP specification and instead provides text for direct inclusion in USAC specification. Other segments of proposed text adds additional explanatory text. The rationale for adding the text is to

Define terms

• 
  Algebraic codebook
  
• 
  Algebraic Vector Quantizer (AVQ)
  
• 
  Closed loop pitch
  
• 
  Fractional pitch
  
• 
  LP coefficients
  
• 
  Zero imput response (ZIR)
  

Add additional descriptive text

• 
  LPD pitch delay (i.e. lag) decoding
  

Removing references to 3GPP specification

• 
  ACELP codebooks
  
• 
  Decoding of AVQ indices
  
• 
  Pitch tracking and gain recalculation in context of bass-postfilter: clarify how the gain of the bass-postfilter is calculated.
  

The presenter noted that all proposed changes affect only the USAC text, and do not impact the USAC reference software.

David Virette, Huawei, noted that the proposed text for decoding of AVQ indices, used the word “encoder.” The presenter clarified that the word “encoder” is used in the context of what is put in the bitstream.

It was the consensus of the Audio Subgroup to adopt into the Study on DIS all changes proposed in this contribution.