Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: USAC vs. HE- and LC-AAC (Read 12626 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.


USAC vs. HE- and LC-AAC

Reply #2
Thanks

USAC vs. HE- and LC-AAC

Reply #3
Will USAC be better than HE-AAC?

At low bit-rates, yes, as Igor has shown. At "ripping bit-rates" of > 96 kbps (no SBR), only on very few audio samples.

Chris
If I don't reply to your reply, it means I agree with you.

USAC vs. HE- and LC-AAC

Reply #4
Will USAC be better than HE-AAC?

At low bit-rates, yes, as Igor has shown. At "ripping bit-rates" of > 96 kbps (no SBR), only on very few audio samples.

Chris


Thanks for the answer Chris.
Can you give a rough evaluation already how far the improvements will go, e.g how good the codec will compete with AAC around 96 or 128 kbps in terms of sound quality?

 

USAC vs. HE- and LC-AAC

Reply #5
Thanks for the answer Chris.
Can you give a rough evaluation already how far the improvements will go, e.g how good the codec will compete with AAC around 96 or 128 kbps in terms of sound quality?

Assuming carefully tuned encoders, at 128 kbps or so, USAC w/o SBR will give about the same average quality as AAC-LC. My guess is that some killer items will sound a bit better (~ 5 MUSHRA points), and none will sound worse with USAC (guaranteed since USAC transform coding is very similar to AAC).

Chris
If I don't reply to your reply, it means I agree with you.

USAC vs. HE- and LC-AAC

Reply #6
Thanks for the answer Chris.
Can you give a rough evaluation already how far the improvements will go, e.g how good the codec will compete with AAC around 96 or 128 kbps in terms of sound quality?

Assuming carefully tuned encoders, at 128 kbps or so, USAC w/o SBR will give about the same average quality as AAC-LC. My guess is that some killer items will sound a bit better (~ 5 MUSHRA points), and none will sound worse with USAC (guaranteed since USAC transform coding is very similar to AAC).

Chris


Sorry Chris, I put my question wrong. I actually wanted to ask how good USAC will compete at low bitrates (like 64 kbps) with SBR against AAC-LC at 96 or 128 kpbs.

USAC vs. HE- and LC-AAC

Reply #7
USAC with SBR at 64 kbps is equivalent to HE-AAC ~70 kbps (max 75 kbps). LC-AAC at 80 kbps is superior than both of them.

USAC vs. HE- and LC-AAC

Reply #8
Thanks again Igor, isn't the development still running, though?

USAC vs. HE- and LC-AAC

Reply #9
As far as I know USAC has reached the final state as format/standard (at last it's final draft).  Later there can be different implementations/encoders of it (those will be written  by different companies). Maybe quality will be optimized later.

Also it should be clear that USAC uses improved but still parametric bandwidth extension (enhanced SBR: eSBR) at bitrates up to 80 kbps.  It means USAC is/should be/will be better than HE-AAC but parametric tools like SBR (even enhanced version of it ) loose completely their advantage  over LC-AAC  and USAC without SBR at bitrates  >~80-85 kbps. 

USAC brings bitrate reduction of about 10% (maybe a bit more or a bit less. Will depend of impelementations, material etc.) at range of bitrates 64 kbps and higher. The efficiency gain is considerably higher than that for lower bitrates (less than 64 kbps).

if You are interested there are some papers about USAC in the web. May the Google be with you 

P.S. The main goal of USAC wasn't to bring much superior compression gain but provide equally high quality for both speech and music.  AMR-WB+ is good for speech and not so much for music and vice versa  AAC/HE-AAC is good for music and not so much for speech. USAC is pretty like enhaced AAC/HE-AAC + AMR-WB+.  Though afaik USAC is the frequency domain (FD) encoder like AAC for ~48 kbps and higher because AMR-WB+ is useful only at less than 48 kbps.

USAC vs. HE- and LC-AAC

Reply #10
Sorry, but I have to object here.

As far as I know USAC has reached the final state as format/standard (at last it's final draft).

USAC will hopefully (if all participants agree) be made an ISO/IEC International Standard at the upcoming 99th MPEG Meeting next month. So development is finished, but there might still be some editorial corrections to the standard document.

Quote
Also it should be clear that USAC uses improved but still parametric bandwidth extension (enhanced SBR: eSBR) at bitrates up to 80 kbps.  It means USAC is/should be/will be better than HE-AAC but parametric tools like SBR (even enhanced version of it ) loose completely their advantage over LC-AAC and USAC without SBR at bitrates  >~80-85 kbps.

This depends on the test set and encoder. When we did internal tests to prepare for the verification tests, we compared USAC with and without SBR at 96 kbps. They were on par. This is possible because due to increased core-coder efficiency (as you mentioned) a higher SBR cross-over frequency can be chosen at higher bit-rates. For some items (e.g. some speech or tonal items without transients) SBR is better than no SBR even at 80-85 kbps.

Quote
USAC brings bitrate reduction of about 10% (maybe a bit more or a bit less. Will depend of impelementations, material etc.) at range of bitrates 64 kbps and higher. The efficiency gain is considerably higher than that for lower bitrates (less than 64 kbps).

Correct. A more accurate rule of thumb for music and mixed content (not for pure speech) is: USAC at x kbps sounds as good as HE-AAC at x+8 kbps. (Edit: looking at the report Igor linked to, this seems to be roughly true for x >= 16.)

Quote
The main goal of USAC wasn't to bring much superior compression gain but provide equally high quality for both speech and music.

Not really. The goal was to sound at least as good as the better of HE-AAC and AMR-WB+ on a per-item basis. And to win am MPEG call for proposals - as the case for USAC in 2008 - superior compression is quite inevitable. And achieving superior performance (especially for stereo or very-low-bit-rate signals) is what we've been working on since 2008. I can safely say that with the given approach (core-coder + parametric stereo and bandwith extension) you cannot get much better quality than what USAC offers.

Chris
If I don't reply to your reply, it means I agree with you.

USAC vs. HE- and LC-AAC

Reply #11
When we did internal tests to prepare for the verification tests, we compared USAC with and without SBR at 96 kbps. They were on par. This is possible because due to increased core-coder efficiency (as you mentioned) a higher SBR cross-over frequency can be chosen at higher bit-rates. For some items (e.g. some speech or tonal items without transients) SBR is better than no SBR even at 80-85 kbps.

Sure, SBR can be better than no-SBR  for a wide range of samples at 80-85 kbps and even at 96 kbps. The question is what happens on average.

It might turn that USAC's SBR and no-SBR are on par at 96 kbps because core coder (without SBR) already does a good job and handle the frequencies up to 16 kHz pretty good. We both know how it's hard to test two codecs that already preserves 16 kHz range well. Add to it error bars.

I'm pretty sure that USAC's SBR and no-SBR are on par at lower bitrates as well, let's say 80 kbps. (in the same test conditions)   

USAC vs. HE- and LC-AAC

Reply #12
Thanks Igor and Chris for the huge amount of information
Btw, is USAC planned as some kind of official successor of AAC, or will it be just an advanced codec for special usage?

Edit:
Sorry, didn't see the split!

USAC vs. HE- and LC-AAC

Reply #13
Thanks for the split!

I'm pretty sure that USAC's SBR and no-SBR are on par at lower bitrates as well, let's say 80 kbps. (in the same test conditions)   

I'm sure they are not  We'll see once we've tuned our USAC encoder to the level of the HE-AAC encoder in, say, Winamp.

Gainless, to answer your question: take a look at this page  or, if you want/can, talk to us at the Mobile World Congress next month.

Chris
If I don't reply to your reply, it means I agree with you.

USAC vs. HE- and LC-AAC

Reply #14
I'm pretty sure that USAC's SBR and no-SBR are on par at lower bitrates as well, let's say 80 kbps. (in the same test conditions)   

I'm sure they are not  We'll see once we've tuned our USAC encoder to the level of the HE-AAC encoder in, say, Winamp.

OK, let put this way. I'm not "pretty sure". I _am_ sure