Talk:Acoustic fingerprint

Vis a vis Hash Code
I removed the header mention of a hash code. A hash code is computed from a series of bits and varies dramatically if even one bit is altered. If an MP3 and WMA file have encoded the same waveform they will still have completely different hash codes. There will be no way to compare the two and know that they represent the same waveform. This would even be true if you just flipped one low order bit in an MP3 frame of one file.


 * Not correct -> a hash function is that you represent something larger / arbitray lenght data into a smaller (mostly) number value. So a method to change bigger object / sound into a number. Quite common in searching
 * if you want also other things like verify stuff, you need a special version of a hash, a cryptographic hash. That type of hash is made to change completely with let's say a single bit different.


 * Currently the most common fingerprints work with direct hashing. With that I mean that if the hash is different the song won't be found. To enhance detection-rate some technologies try to predict with bit's could be wrong and also search those. — Preceding unsigned comment added by 83.161.136.61 (talk) 07:41, 19 September 2014 (UTC)

In contrast, an acoustic fingerprint is computed from the rendered waveform. In the case above the waveform rendered from the MP3 and WMA file will be different, but will have similar acoustic fingerprints. Or at least an acoustic fingerprint technology would be able to compare the two and know that they are from the same waveform. This is very different from a hash code of a file or a digital representation of a waveform.

I left in the mention of deterministic, because the algorithms don't typically involve randomness. So strictly speaking the algorithms are deterministic. However, I think the article would be more clear if we also removed that. Unless I see discussion against it, I'll remove the word in a few weeks. Jimbo (talk) 23:41, 4 October 2008 (UTC)

Anti acoustic fingerprinting (speculative)
The hackers of the world need to come up with robust anti acoustic fingerprinting technology. This should basically involve adding random noise without reducing listening quality. Theoretically, it should be easy to come up with. Any info on this topic will be appreciated as it becomes available in the future. --Amit 04:45, 13 October 2006 (UTC)


 * Acoustic fingerprints are generally designed to analyze audio in the same general way that humans perceive it; The worst anyone could probably pull off is to make fingerprints not match with the already existing and "clean" recordings - but the tampered copies would still be identified as one. I would say it's also economically infeasible to separately tamper each copy of a recording, with different seeds – unless it's produced in very low quantities. In short, I'm sceptical about the effectiveness of this, not to mention that I fail to see any practical economical use for such a technology. -- intgr 05:50, 13 October 2006 (UTC)


 * But there is a potential practical use — although not an economical one as I can tell. YouTube under Google is planning to implement (as some other sites already have) acoustic fingerprints supplied by media companies to weed out or disallow material. Robust anti acoustic fingerprinting could allow not only for user freedom, but also safeguard YouTube's popularity.


 * Whether a tampered copy would still be matched by a fingerprint for the original or not depends upon the robustness of the algorithms used on both sides. --Amit 06:56, 13 October 2006 (UTC)


 * Good point, I never thought of fingerprints being used for these kind of things. As for the latter, I just intended to state the obvious that copies of a single tampered file will still match a single fingerprint, even if it's different from the original. :) -- intgr 11:04, 13 October 2006 (UTC)

Almost all fingerprints can be easily fooled by pitching. Which sounds almost the same for humans and changes everything for fingerprints. Adding noise has no effect. — Preceding unsigned comment added by 83.161.136.61 (talk) 07:50, 19 September 2014 (UTC)

Image and video fingerprints
There exist image and also video fingerprint technologies, and these deserve mention on Wikipedia. At least two options exist:
 * 1) Have a separate article for each type of item.
 * 2) Move this article to Media fingerprint.
 * 3) Separate specifics of fingerprinting each type of item, while merging commonalities.
 * 1) Separate specifics of fingerprinting each type of item, while merging commonalities.

Opinions are needed! --Amit 07:10, 13 October 2006 (UTC)


 * This particular article seems pretty specific to acoustic fingerprinting – I think you would be better off starting an entirely new article for media fingerprints and linking to more specific articles from there. -- intgr 11:04, 13 October 2006 (UTC)


 * Agreed. Jimbo (talk) 23:43, 4 October 2008 (UTC)

Removed from before list
I removed this from before the list because it seemed like it was instructions for writers and didn't belong in the article: "Products on this list should either have an accompanying existing article link which verifies their notability for acoustic fingerprinting, or reliable sources as footnotes against the name showing they are notable for this reason." I'm sure it is useful to at least mention it here, though. Cleanelephant (talk) 08:59, 26 December 2011 (UTC)

merge
Isn't a spectrogram a (more elaborate) version of a simple fingerprint? then there would be no difference in the statements between acoustic fingerprint and search by sound. -- Kku 13:58, 25 February 2016 (UTC)


 * An acoustic fingerprint starts with a spectrogram. But just like a human fingerprint starts with an image, there's a lot more that goes into making the identification. ~Kvng (talk) 14:55, 11 August 2017 (UTC)

Search by sound merge proposal
Merge banners have been up for this proposal since 2015 or 2016 but I see no discussion. I support this proposal and will do it soon unless there are objections here. ~Kvng (talk) 14:53, 11 August 2017 (UTC)

I don't think the merge is a good idea. I think Acoustic Fingerprinting is a specific topic that could use more information here. Reading the Search By Sound page I find the part on fingerprinting reads like an ad for Shazam, not as an authoritative article on the subject. I could agree to reducing the Search By Sound page to something much lighter that references other pages on techniques that can be used to achieve that objective. Jimbo (talk) —Preceding undated comment added 16:20, 11 August 2017 (UTC)


 * I'm having trouble understanding your thinking here. The proposed merge would add more information to this article which is what you're asking for. WP:PROMOTIONAL issues can be addressed at any time and I don't see how they factor into a merge decision. ~Kvng (talk) 16:34, 29 September 2017 (UTC)

An acoustic fingerprint is comprised of a set of measurements of the audio signal taken over some time frame. Search By Sound is one application that might use a type of Acoustic Fingerprint to achieve the query objective. If we merge SbS into this, then we should probably add other examples of the use of acoustic fingerprints. Jimbo (talk)jschrempp


 * That would be a welcome improvement but one that can be separate from the merge proposal. What other applications for acoustic fingerprint are you aware of?
 * I'm still not clear whether there's an objection to this merge. I'm going to start by moving material about acoustic fingerprint from Search by sound to this article. That seems like an uncontroversial first step. ~Kvng (talk) 15:08, 1 October 2017 (UTC)

Other uses include: speaker identification (perhaps in an archive), copyright compliance for performance rights (where exact match to a specific performance of the work is critical and false positives are completely unacceptable), sound similarity (searching a sound effects database for example), duplicate detection (for instance in a library). These can all use acoustic fingerprints but with far different objectives than "search by sound". Jimbo (talk)


 * I shuffled some material on 1 October 2017‎. It does not look like there is consensus to merge at this time. I have taken down the merge banners. ~Kvng (talk) 16:07, 18 November 2017 (UTC)

exact match
Note that while it does a fuzzy match to allow for noise, Shazam does an exact match to the CD track. That is, it won't match a similar recording, even by the same artist, at least as well as I know it. It might be nice to have one that did. Gah4 (talk) 10:02, 15 April 2021 (UTC)
 * , see Search by sound. ~Kvng (talk) 13:43, 18 April 2021 (UTC)

Shazam bias?
There's a whole section on Shazam's internal functionality, shouldn't this go in Shazam's own article? Shouldn't other applications be discussed? Seems biased that it's the only software mentioned in this article and has an entire section dedicated to it. Halogen487 (talk) 17:13, 9 October 2022 (UTC)


 * I assume we got here because Shazam was the first to offer this technology commercially. It is a reasonable suggestion to move the Shazam section content to Shazam (application) but that would leave us without a technical explanation here of how the magic works. ~Kvng (talk) 16:47, 15 October 2022 (UTC)