Experimental condition
|
Compression gain
(FD frames only)
|
8-bit PCM global gain, arithmetic coding of scalefactors, separate arithmetic coding of spectrum
|
0.62 %
|
8-bit PCM global gain, joint arithmetic coding of scalefactors and spectrum
|
0.71 %
|
joint arithmetic coding of differential global gain, scalefactors and spectrum
|
1.03 %
|
Ralf Geiger, FhG, presented
m16685
|
Considerations on Coding of Critical Bitstream Elements
|
Ralf Geiger, Ralph Sperschneider, Markus Multrus
|
The contribution observes that global gain, as an 8-bit PCM value, permits very simple level adjustment. It further observes that jointly coding scalefactors and spectum in a single arithmetic coding token means that the token “crosses” syntax element boundaries, and also does not permits scalefactors and spectrum to be put in different error sensitivity categories.
Henley Oh, LG, noted that there is no global gain in ACELP or TCX, or if there is, its stepsize if much greater than what is provided in 8-bit global gain. Hence, in USAC, there is currently no unified way to adjust global gain.
The Chair encourages experts to investigate:
-
a method for adjusting the level of the entire USAC decoded wavform (i.e. FD, ACELP and TCX frames)
-
whether there are practical applications for such level adjustment
As a separate but related issue, the Chair requests that experts in the group should clarify how error resilience might be relevant in CE evaluation.
Werner Oomen, Philips, presented
The contribution notes that USAC currently uses parametric coding of stereo at low bit rates and M/S, L/R coding at higher rates. It proposes to integrate these two methods by using parametric stereo coding with a residual signal. At the lowest bit rates, no residual is used, at intermediate rates, a bandlimited residual is used, and at highest bitrates a full-band residual may is used. The “downmix” and “residual” signals are treated as a regular monophonic coder channel. The contribution has been evaluated at 64 kb/s, which implies that only FD mode can be used.
Listening test results were presented for 12 subjects, 15 CfP test items and 2 additional items. The test results show a improvement in average performance of the absolute MUSHRA scores that is statistically significant at the 95% level of significance for all items together and for five individual items (3 from the CfP set).
The proponent requests that interested experts help draft a workplan and interested companies assist in a performing a cross-check. It is anticipated that the CE proponents will profile all components of CE proposal at the next meeting.
Max Neuendorf, FhG, presented
m16579
|
Proposed Corrections to Reference Software and WD3 of USAC
|
Max Neuendorf
|
The contribution proposes a number of editorial changes to WD3. It was the consensus of the Audio Subgroup to adopt these changes into a WD4 to issue at this meeting.
Werner Oomen, Philips, presented
m16666
|
Consideration on efficiency of USAC syntax elements
|
Bernhard Grill, Kristofer Kjorling, Werner Oomen, Philippe Gournay
|
The contribution notes that there are a number of CEs that whose performance might be modest. It asserts that adoption of these tools early in the standardization process might unduly impact the flexibility of the codec architecture. It reports on an investigation of two tools that are currently in USAC, that appear to offer only modest increase in compression performance, but that present difficulties in decoding architecture. These are
-
Time slot position coding
-
Pilot coding
The contribution proposes to move these tools to “Syntax Elements under Consideration” for assessment no later that when the standard progresses to CD stage.
The Chair noted that this is a first step in an ongoing process of evaluating tools in USAC to verify that all bring a meaningful level of performance.
There was considerable discussion on the proposal. The proposal will continue to be discussed during the week.
Dostları ilə paylaş: |