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: Would a big ITU-R BS.1387 Analysis be interesting? (Read 5907 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Would a big ITU-R BS.1387 Analysis be interesting?

I was thinking about if it would have any interest to do a big test on a lot of real music with that new ITU-R BS.1387 Analysis tool?

I am willing to put some time into doing it if ppl would be interested in such a analysis.

My intention was to test mpc -standard, -xtreme, -insane, the alt presets, r3mix, and some ogg (which setting should be tested here?). Any other setting that would be interesting? (if the test is done)



Regards,
Jan.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #1 –
Remember that this tool is not a replacement for a good listening test. It can be very helpful in phases of codec development where listening tests are impossible, or to be integrated into automatic QoS measurement system (in broadcasting) - but using it as a measure of absolute codec quality could be misleading.

I can confirm that it misses some pre-echo events, and even marks some clips significantly better than other where the "real world" results are different! Therefore it can't be used as "proof" that some codec sounds better than other.

I am saying this just because we'll have a whole bunch of newbies claiming that something sounds better where it doesn't - like those people that do Cool Edit spectral analysis and then come back saying that some codec is the best because of "flat" frequency response

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #2 –
Has this link been put up before?:

http://home.wanadoo.nl/~w.speek/eaqual.htm
Ruse
____________________________
Don't let the uncertainty turn you around,
Go out and make a joyful sound.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #3 –
IMO we should first find out how accurate the results from EAQUAL are. It would be nice if some people with lots of listening experience would compare the results of EAQUAL on some test samples with what they hear theirselves.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #4 –
Ok, allright - here is one sample that especially bugs EAQUAL:


Sample #1: http://www.psytel-research.co.yu/samples/cast1.aac
Sample #2: http://www.psytel-research.co.yu/samples/cast2.aac

Offset for cast1.aac should be 1172, I think
and for cast2.aac offset should be 1024

(with latest FAAD)

For cast1.aac ODG is -0.73
For cast2.aac ODG is -0.52

Now, according to EAQUAL, cast2.aac should sound better than cast1.aac, or at least equal to - BUT..

try listening to these and you'll notice that cast2.aac suffers from drop-outs in first 4 hits, and bad fast castanets hits... clearly worse than cast1.aac!

Cast1.aac was encoded in MAIN profile AAC with the reference AAC encoder, similar to the version used in MPEG listening tests but improved even more.

Cast2.aac was encoded in LC profile AAC with another commercial encoder.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #5 –
I measured the offset of cast1 (decoded with latest FAAD) and think it should be 1203.

With that offset the results are nearly equal:

ODG cast1 = -0.5991
ODG cast2 = -0.5265

Of course I listened to both aac files and compared them to the original. I'm sorry to say that in my ears they are both transparant :-(

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #6 –
I'm not asking to do the final test - I'm just asking if an analysis of a lot of material will have any value or I would just be wasting my time.


Jan.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #7 –
Well, it depends - we have already proven that PEAQ "Basic" model misses some temporal distortions - I don't know, it would be nice to see some initial results on mainstream codecs, but...

It might be a good idea to wait for advanced version of EAQUAL with ITU BS-1387 "Advanced" model (If Alexander is going to implement it). In Opticom OPERA product sheet, as well as in AES paper covering BS-1387 it is stated that "Advanced" model provides much better temporal resolution than "Basic" version.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #8 –
Ok, better wait then. 


Jan.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #9 –
Where can I read more about EAQUAL?

I gather that the "basic" version is implemented, and that the "advanced" version may be implemented in the future. Where can I read about how the basic version was implemented? There are various aspects of the published standard which are open to interpretation, and I'd be interested to see how the mapping from internal variables to ODG is accomplised in the code on Sourceforge.

I know the obvious answer is "go and read the code" - but I'm not a programmer! Where can I find more information?

Cheers,
David.
P.S. - Happy new year (since it's my first post of 2002).
P.P.S. - Don't email me - my mailbox exceeded 70MB over the Christmas holiday, and shut down.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #11 –
Quote
Originally posted by 2Bdecided
Where can I read more about EAQUAL?

I gather that the "basic" version is implemented, and that the "advanced" version may be implemented in the future. Where can I read about how the basic version was implemented? There are various aspects of the published standard which are open to interpretation, and I'd be interested to see how the mapping from internal variables to ODG is accomplised in the code on Sourceforge.

I know the obvious answer is "go and read the code" - but I'm not a programmer! Where can I find more information?

Cheers,
David.
P.S. - Happy new year (since it's my first post of 2002).
P.P.S. - Don't email me - my mailbox exceeded 70MB over the Christmas holiday, and shut down.


David, I suggest you to get and read following papers:

AES #4931 - "A Combined Measurement Tool for the Objective, Perceptual based evaluation of Compressed Speech and Audio Signals"  (M. Keyhl, Ch. Schmidmer, H. Wachter, Opticom GmbH, Erlangen Germany)

and

AES #4512 - "Objective Audio Signal Evaluation - Applied Psychoacoustics for Modeling  Perceived Quality of Digital Audio" (Th. Sporer, University of Erlangen)

While the first one is more or less a white paper for OPERA software, second one is more interesting from the scientific point of view.

I also highly recommend K. Brandenburg and M. Kahrs' book  "Applications of Digital Signal Processing to Audio and Acoustics" (First Chapter)

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #12 –
The price is crazy!!! (150$)

Is this book really worth the price?

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #13 –
Well... it has 450 pages  However, only first two chapters are related to audio coding and perceptual evaluation of audio quality - rest of the book is related to DSP application in other audio research topic (click reduction, hearing aid tool design, wavetable synthesis design, etc..) - it is a good DSP book, not related only to audiocoding.

Karl's chapter is very interesting, but already seen in many of his papers - he describes vital parts of psychoacoustic model with information how to measure tonality, various methods for pre-echo protection, future of audio coding... and describer MP3 and AAC codec tools.

PEAQ chapter from Beerends is also interesting, and one could learn about psychoacoustical model used in KPN and Opticom tools.

Not sure is it worth 150$, though.

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #14 –
Quote
Originally posted by 2Bdecided
Where can I read more about EAQUAL?

I gather that the "basic" version is implemented, and that the "advanced" version may be implemented in the future. Where can I read about how the basic version was implemented? There are various aspects of the published standard which are open to interpretation, and I'd be interested to see how the mapping from internal variables to ODG is accomplised in the code on Sourceforge.


The standard is a quite, good explanation of the algorithm, especially if you don't try to implement it :-)

If you have detailed questions, you may ask me directly.

Before I begin to post tons of references, how deep do want to go in this stuff?
Some references are mentioned at http://www.peaq.org, but I think they already have been named.

Alexander
 

Would a big ITU-R BS.1387 Analysis be interesting?

Reply #15 –
Quote
Originally posted by Ivan Dimkovic
It might be a good idea to wait for advanced version of EAQUAL with ITU BS-1387 "Advanced" model (If Alexander is going to implement it).


I see, you want to get me to work :-)

Well, this is quite much work. Perhaps somebody of you wants to help?

Alexander