2

I just ran into https://skinmotion.com/ which sells "Soundwave Tattoos".

enter image description here

My gut says tattoos cannot have high enough resolution to actually encode 15s of audio - even at speech quality.

Reading all the "FAQ" items I get the impression that the company knows the tattoos are "unplayable" except with their specific App.

This all suggests that their app merely detects the shape, making a fingerprint and looking up matching samples in their cloud database. The match can be crude, as long as there is sufficient confidence gap to distinguish from (expectable) similar matches.

I have two questions:

  1. has anyone done the maths to rule out/confirm that tattoos cannot actually encode the required data volume

  2. does anyone know how to approximate at what database size (number of registered, activated tattoos) there would be, say, a threshold probability p (say, 10%) of the app playing the wrong sample when scanning any particular tattoo?

I'm aware many parameters are highly implementation-dependent, and likely patented information. However, I also know that mathematics has general formulae about information representations, entropy and compression vs. information loss. So I figure mathematically there should be an upper bound on the performance of a service like this.

sehe
  • 123
  • 4
  • 2
    After looking at it, I don't see where they make the claim that you're implying they make. The FAQ explicitly states it works by "A combination of audio processing, **image recognition**, computer vision, and **cloud computing** to create a mixed reality experience", "We generate the shape of Soundwave from that file which **you can trim the beginning and end off** and change the height and width", and "the Skin Motion app will only play back Soundwave Tattoo if the user is logged in to their account and the Soundwave Tattoo is created and **activated** by that user" – paradisi Jan 04 '18 at 00:17
  • @sumelic I'm not saying they're claiming it. They're selling it as-if though (basically, you need to trust their service to stay around forever, as tattoos don't go away). I was asking whether it is in principle feasible to create an independent app to play back the sample, at useful fidelity (question 1.) – sehe Jan 04 '18 at 00:38
  • 1
    By the way, thanks for highlighting those passages of the FAQ that indeed informed my expectations. And for my suspicions about viral marketing/emotional appeal: https://twitter.com/sakyrahhh/status/948381469650464768 – sehe Jan 04 '18 at 00:40
  • 1
    I can understand why you'd be interested in learning the answer to that question, but I'm not sure it's a great fit for this site. The kind of "experts" who write answers on this site are mainly expert at looking up established sources about [notable claims](https://skeptics.meta.stackexchange.com/questions/2506/what-is-a-notable-claim). This site doesn't necessarily select for experts in any particular areas of science. For that reason, answers based on [personal research are discouraged](https://skeptics.meta.stackexchange.com/questions/2924/faq-what-constitutes-original-research). – paradisi Jan 04 '18 at 00:42
  • Mmm. It's the first time I use this site, so unsurprisingly my expectations may have been wrong. Thanks anyways. I'll get the hang of that more, I suppose. – sehe Jan 04 '18 at 00:44
  • 1
    No worries! I think it might be better for you to ask this somewhere else where you can be more confident that you'll be getting an answer/answers from people who actually know the relevant science. To me, that seems likely to be phonetics and computational linguistics; sadly, I'm not sure if the Linguistics.SE site would welcome a question like this (the people there can be a bit unfriendly about unconventional questions), but perhaps you'd get a good answer on Physics.SE. – paradisi Jan 04 '18 at 00:44
  • Let us [continue this discussion in chat](http://chat.stackexchange.com/rooms/71156/discussion-between-sehe-and-sumelic). – sehe Jan 04 '18 at 00:44
  • 1
    I've closed this, because no-one is making the claim. It seems far more likely that it is a database lookup rather than scanning the waveform. The question about whether it could be done is speculative, and off-topic here. – Oddthinking Jan 04 '18 at 02:35
  • 3
    @sehe: Feel free to repost the question "Does a Tatoo have enough resolution to encode an audio wave form" and physics or signal processing stack exchange. The answer will be "no" – Hilmar Jan 05 '18 at 13:52
  • The linguist Mark Liberman has made a blog post that seems to support your own thoughts about this topic: http://languagelog.ldc.upenn.edu/nll/?p=36152 – paradisi Jan 06 '18 at 02:19
  • That's an awesome find, @sumelic Thanks a bunch! – sehe Jan 06 '18 at 02:19
  • "is in principle feasible to create an independent app to play back the sample" - no, because the tattoo doesn't contain the samples. – Mazura Jan 12 '18 at 05:41
  • @Mazura , there's and android app that tries to read sound waves like in the image above. App is called PhonoPaper. It's complete garbage from my experience. https://play.google.com/store/apps/details?id=nightradio.phonopaper&hl=en_GB – LateralTerminal Jan 12 '18 at 18:43

0 Answers0