[M4IF Technotes] Re: MPEG-4 AAC-LC: Buggy Conformance Streams

[Ralph Sperschneider]

Umang Garg wrote:
> Dear Members, Hello,
> 
> 
> I understand that the MPEG-4 AAC-LC conformance streams are available at 
> the following FTP site:
> 
> ftp://mpaudconf:adif2mp4@ftp.iis.fhg.de/mpeg4audio-conformance
> 
> I currently have with me an ISO provided 'reference software' that 
> contains the MPEG-4 AAC Decoder ( version 2 ).  We purchased that code 1 
> month back from ISO.
> 
> In the reference software the AAC decoder is available along with all 
> its tools within the following directory structure: 
>  /audio/natural/rewrite/mp4AudVm
> 
> 
> Now when I try to pass the interim conformance streams available from 
> the FhG, I find them to be buggy !!!
> 
> 
> For Example:
> 
> Bitstream number al03_44.wav is a reference wave files with a single 
> channel ( channel 0 ) .
> 
> al03_44.mp4 is its corresposing reference encoded stream.
> 
> When I pass al03_44.mp4 through the reference decoder available from the 
> MPEG ( ISO/IEC), I get an output wave file which has two channels ( 
> channel 0 and channel 1) !!! This is just one of the many examples that 
> I can give.
> 
> 
> Now obviously this means that I cannot establish the conformance test 
> criterion for the generated .wav output with respect to the reference 
> .wav file
> 
> 
> Even more interestingly, FhG seems to have come up with their own series 
> of AL files starting from al14_*.* to al19_*.*
> 
> None of these files find a mention in the MPEG-4, Part 4, Conformance 
> Testing document.
> 
> If what I have written above makes sense then it is really a catch-22 
> situation. The MPEG-4 AAC Decoder cannot be validated until some body ( 
> hopefully ISO ) comes up with a set of correct conformance test streams.
> 
> What I would really like to know is:  Is this a genuine problem ?  Or am 
> I doing something terribly wrong ....
> 
> Thanks,
> 
> Umang Garg
> NeoMagic Design Center
> 

Dear Umang,
please find below some comments with regard to your observations.
reference software:
-------------------
I can confirm the behavior you describe (just downloaded the code from 
http://www.iso.ch/iso/en/ittf/PubliclyAvailableStandards/ISO_IEC_14496-5_2001_Software_Reference/ 
and compiled it), but have no idea why that decoder generates an output file 
with two channels while decoding al03_44.mp4. As long as you only like to deal 
with AAC (Main, LC, SSR or LTP), you might alternatively use the multichannel 
AAC decoder (available under /audio/natural/refSoft/mp4mcDec). I checked this 
decoder as well and it produced just one channel. Thus, the sequence is correct, 
but the rewrite decoder is malfunctioning.
Anyway, you should consider that the software you purchased from ISO is outdated 
(it is about 2 years old). It always takes ISO a significant time to publish any 
new standard piece. So one reason to become an MPEG member might be to have 
access to weekly snapshots of the MPEG-4 audio reference software. With the 
current rewrite decoder I get just one channel if I decode al03_44.mp4 (hence, 
this bug has been fixed meanwhile ;-)).
conformance test sequences:
---------------------------
All al* conformance test sequences on our server (except of al15, which is 
currently in the process to be updated) have successfully been cross-checked. 
All of them can be decoded with the multichannel decoder. The rewrite decoder 
does not yet support DRC and multichannel decoding (just mono and stereo).
It is not that FhG has come up with their own series of test sequences. The 
sequences on our server are even not the property of FhG. We just support the 
work of MPEG on a complimentary basis in hosting these sequences and furthermore 
in providing access to them even for non-MPEG members.
* al14, al15, al16 and al17 have been added by means of ISO/IEC
   14496-4:2000/Cor.2 (unfortunately not yet published):
   - al14, al15 and al16 test DRC (previously there where only sequences for AAC
     Main available to test DRC).
   - al17 tests the evaluation of the element_instance_tag (it provides two SCEs,
     where the order of them switches from frame to frame)
* al09 and al10 have been renamed to al18 and al19 to avoid name conflicts with
   MPEG-2 AAC conformance test sequences.
All these issues are covered by the second edition of ISO/IEC14496-3, which is 
also not yet published (again, it is out of MPEG´s hands, but needs to be 
processed by ITTF).
I hope this helps for the moment. I know that its not always easy to understand 
what is going on in MPEG, especially since it takes occasionally a longer time 
for ISO/ITTF to publish new standards. However, MPEG members are usually nice 
and answer any questions related to their standardization work, as long as they 
are asked in a friendly manner ;-).
Best regards,
Ralph
-- 
Dipl.-Ing. Ralph Sperschneider  | Phone: +49 9131 776 344
FhG IIS                         | Fax:   +49 9131 776 398
Am Wolfsmantel 33               | mailto:sps   iis.fhg.de
D 91058 Erlangen                | http://www.iis.fhg.de/amm/