IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
FLAC made from iTunes AAC LC - how to detect?
EpicForever
post Jan 18 2013, 19:22
Post #1





Group: Members
Posts: 651
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



Hello.

Since about 3 years I buy music at Beatport and Junodownload, sometimes at Trackitdown. I prefer CD quality, so mostly I buy lossless files (WAV/AIFF at Beatport and FLAC at Juno). Unfortunately several times I found that some of files that I bought were just transcodes from some kind of lossy format. I assume that were transcodes from MP3, as it's most popular and bandwidth of audio signal was cut off at specific frequencies like 16,5 kHz (128kbps), 18,6 kHz (192kbps) or 20,1 (320kbps) kHz. It is easy to find such thing - I can just see spectrogram in foobar2k (with linear frequency scale selected) or use auCDtectTaskManager. But about month ago I bought one album at iTunes - just because of iTunes bonus track available only with full album. I investigated how it is encoded (I've never tried AAC at higher bitrates before) and I found that iTunes albums have full frequency scale, despite they are lossy... There's no low pass filtering known to me from MP3. I've transcoded whole album into FLAC with foobar and then I passed FLACs through auCDtectTaskManager. All files were concluded as 100% CDDA... I was shocked. So my question is - is there any other method to determine authenticity of lossless file? Any software tool that uses better algorithm than auCDtect, or just different algorithm tuned specifically for detecting HQ AAC transcodes?

Thank you in advance for any help.
Go to the top of the page
+Quote Post
saratoga
post Jan 18 2013, 19:34
Post #2





Group: Members
Posts: 4715
Joined: 2-September 02
Member No.: 3264



QUOTE (EpicForever @ Jan 18 2013, 13:22) *
So my question is - is there any other method to determine authenticity of lossless file? Any software tool that uses better algorithm than auCDtect, or just different algorithm tuned specifically for detecting HQ AAC transcodes?


There is not. And just because something is lowpass filtered doesn't mean its from lossy. The reverse is also true, one can easily make MP3 files that are not lowpass filtered.
Go to the top of the page
+Quote Post
EpicForever
post Jan 18 2013, 21:47
Post #3





Group: Members
Posts: 651
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



I know. But lowpass filtering used as element of production/mastering is mostly gradual - slopes aren't steep, transition band is wide. Spectrum peaks on spectrogram just diffuse. But if I see sharp edge on spectrogram, at 16,5 kHz it must be lossy compression artifact... Some cut off frequencies are specific for certain bitrates. Other thing is that I knew tracks that turned out to be transcodes before I bought them - from mixed CDs (holded in my hands) or vinyl rips and I knew that they got full frequency band.
I also tried "-k" switch in LAME 3.96/3.97beta and I know that MP3 can be done without filtering. But I didn't know that frequency response is the only factor used in analysis of losslessness of audio files... Are you sure that there isn't anything else that can be... computed? extracted? from signal that could tell us weather file is lossy or lossless (with some probability of course)?

This post has been edited by EpicForever: Jan 18 2013, 21:48
Go to the top of the page
+Quote Post
Woodinville
post Jan 18 2013, 21:50
Post #4





Group: Members
Posts: 1401
Joined: 9-January 05
From: JJ's office.
Member No.: 18957



If you can find a section of your music with a stationary set of tones, see how deep the noise floor is in an area of signal. Not definitive, though. You really need the original. sad.gif


--------------------
-----
J. D. (jj) Johnston
Go to the top of the page
+Quote Post
Dynamic
post Jan 19 2013, 12:50
Post #5





Group: Members
Posts: 793
Joined: 17-September 06
Member No.: 35307



Sometimes listening to the SIDE channel signal of mid-side encoding (i.e Left - Right sample values), will reveal lossy encoding effects, as can using very clean but extreme EQ to try to de-mask the varying quantization noise at certain frequencies. However, even that isn't definitive, especially with the number of legitimate DSP effects around, some of which will leave signatures that would be fully masked in normal listening (especially de-noise algorithms).

If you think you've found one, I guess you COULD post a short (<30s) lossless sample from a certain time in a certain track and ask for people with the CD version to bit-compare or upload the equivalent section of a CD release for comparison in compliance with the TOS.
Go to the top of the page
+Quote Post
Mach-X
post Jan 20 2013, 03:03
Post #6





Group: Members
Posts: 235
Joined: 29-July 12
Member No.: 101859



QUOTE (Dynamic @ Jan 19 2013, 06:50) *
Sometimes listening to the SIDE channel signal of mid-side encoding (i.e Left - Right sample values), will reveal lossy encoding effects, as can using very clean but extreme EQ to try to de-mask the varying quantization noise at certain frequencies. However, even that isn't definitive, especially with the number of legitimate DSP effects around, some of which will leave signatures that would be fully masked in normal listening (especially de-noise algorithms).

If you think you've found one, I guess you COULD post a short (<30s) lossless sample from a certain time in a certain track and ask for people with the CD version to bit-compare or upload the equivalent section of a CD release for comparison in compliance with the TOS.

What he said play the suspected files through a hafler circuit...left and right positive connected to a single speaker, lossy encoders toss this info out first since its inaudible in regular stereo, if its particularly "swishy" sounding its a lossy source. Eqing can also help but is harder to detect than an isolated hafler speaker. I have long believed some people might be cheating abx tests with these methods!
Go to the top of the page
+Quote Post
EpicForever
post Jan 20 2013, 10:58
Post #7





Group: Members
Posts: 651
Joined: 14-September 11
From: Szczecin, PL
Member No.: 93712



Thank you for your help. I see it's really tricky question... I thought that it is possible to find it using computations, because lossy coding is mathematical operation - thus it could be mathematically detected (with specific values of some advanced signal parameters for example). I see I was completely wrong... Fortunately it's very unlikely to happen, that Beatport, Juno, Audiojelly, Trackitdown or DJDownload will have transcode made from AAC (as it's much rarer format in music industry than MP3). Besides the fact that transcodes in webshops are generally very rare (but unfortunately occurs sometimes). MP3 transcodes are very easy to find, because no one uses specifically tuned parameters in industry scale coding of MP3 files for webshops. Standard presets for LAME (as "-h"/"-q2") always expose transcoding in spectrograms.
Because someone revoked TOS and I think he said about TOS#8 (ABX-ing) I have to explain something. I am not trying to find transcodes because of my "golden ears" (which I don't own - I hear quiet but constant hiss above ca. 14~15kHz). It's all about money. If I pay 1$ more at Beatport for AIFF then I have to get what I paid for - full frequency scale and not flat cut at 18,6 kHz. 192mp3 kbps mp3 can be easily obtained from the Net, without paying for it. If I pay 1,49GBP for 320kbps MP3 then I don't want quality of 128kbps MP3 encoded with "-fast" preset. That's why I asked for sotware tool - objective algorithm based on scientific knowledge. I needed easy to obtain and easy to recreate evidence wheather something is transcoded or not, WITHOUT judging weather it is audible or not.
My conclusion is that the answer for question from the topic is: as for today it is impossible to objectively detect fake lossless file in case it was transcoded from iTunes AAC LC files - currently there's no such software tool/algorithm that could find this.

This post has been edited by EpicForever: Jan 20 2013, 11:01
Go to the top of the page
+Quote Post
db1989
post Jan 20 2013, 16:50
Post #8





Group: Super Moderator
Posts: 5141
Joined: 23-June 06
Member No.: 32180



QUOTE (EpicForever @ Jan 20 2013, 09:58) *
Because someone revoked TOS and I think he said about TOS#8 (ABX-ing) I have to explain something. I am not trying to find transcodes because of my "golden ears" (which I don't own - I hear quiet but constant hiss above ca. 14~15kHz).
I suspect Dynamic’s reference to the TOS was just (A) to note that samples of copyrighted material >30 s are not allowed by TOS #9, and (B) to suggest that others try ABXing, rather than saying that you had said anything controversial. smile.gif

Your reasoning that retailers should not be able to sell ‘lossless’ content that is transcoded from lossy is totally correct, of course!
Go to the top of the page
+Quote Post
Dynamic
post Jan 21 2013, 11:22
Post #9





Group: Members
Posts: 793
Joined: 17-September 06
Member No.: 35307



Yes, I did mean TOS#9 that if you post a sample it should be no more than 30 seconds for copyright reasons.


--------------------
Dynamic – the artist formerly known as DickD
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
2 User(s) are reading this topic (2 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 16th April 2014 - 08:42