I woudn't be sure the difference is as clear as you make it, some recordings, for example CDs that were made from old recordings (the 60s ot the 70s), sometime simply don't contain high frequencies at all, or at a very low level. Similarly, I have some CDs which were recorded at a whooping -35dB average to allow for really high dynamics, and these contain very little HF info too. Do I need to talk about badly mixed pop music for which the samples used were already MP3s at the production stage?
So looking for the missing frequencies is a bad method.
+1 transfers from old tapes/digital tapes for classical sometimes you get a FR from about 100hz to 8khz
... Sometimes whatever digital format they were stored in could well be a variation of Mpeg (l2, l3
) so Cut-offs and inspecting the higher end spectrum for artifacts is a bit hit and miss.
Tau is for CDs, Aucdtect/Audiochecker (and a better AUCDTect GUI http://y-soft.org/English/products/auCDtect-Task-Manager/ ) are for encoded lossless files (basically it's a GUI on top of AUCDTect and WV,FLAC,APE, etc decoders. Decodes to wav, passes the wav to aucdtect and spits out the log.
Aucdtect is not very well regarded. It uses the analysis techniques described in http://en.true-audio.com/Tau_Analyzer_-_Aucdtect_Algorithm_Details
which appear to be sound, but people argue it doesn't spot non mp3 compression too well, especially at higher bitrates (ogg, aac, etc)
It also won't tell you whether someone decided to volume level the entire CD before encoding it
...
Anyway, if anyone has ACM access and can pull http://portal.acm.org/citation.cfm?id=1597828 for us, I'd be interested to have a read (or anything on the similar theme that follow on from it or relates to it) . edit: ooh found it http://www.fileden.com/files/2009/2/14/2321055/My%20Documents/MP3%20Bit%20Rate%20Quality%20Detection%20through%20Frequency.pdf or something like http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=4490000 / http://www.csie.nctu.edu.tw/~cmliu/Courses/Compression/Artifacts.pdf , since the other one is not cited...
(unfortunately there aren't automated tools for the later papers
). UNfortunately, most papers seem to deal with low bitrate artifacts for things like VoIP, or watermarking or forensics...