| Author |
 Replies: 19 / Views: 3,217 |
|
New Member
Belgium
10 Posts |
Hey guys,
I wanted the coincommunity forum regulars to be the first to share this with.
Over the past couple of months I've teamed up with fellow gold-bug bitten engineers and we've been working on a new way of authenticating bullion coins.
In a nutshell: We're able to authenticate gold coins by their acoustic fingerprint.
The age old "ping test" -- where you balance the coin on your finger, gently strike it and listen to its pitch -- is a crude application of this acoustic method.
Listening for a gold coin's high pitched sound, as a method of verifying its gold contents, is likely to be as old as gold coins themselves.
We've set out to update this method to the 21st century. Although humans aren't equipped to accurately distinguish between tones exceeding 5 kHz, computers do a great job at this.
We've built a prototype acoustic coin authenticator which works in your browser. Right now we're supporting 3 coins. All you need is a functioning microphone and at least one 1 oz gold Krugerrand, American Eagle or Maple Leaf.
Are you intrigued, skeptic or amused? We're looking for critical beta testers to help test our system. Interested?
Edited by jessegold
07/31/2013 11:33 am
|
|
|
|
Pillar of the Community
United States
1200 Posts |
1---Welcome to CCF.
2---What you've got sounds definitely interesting.
3---I think posting your email address is a big no-no here and I suspect one of the mods will be along shortly to reinforce that point.
Good luck with your new sonic test thing...
|
|
New Member
 Belgium
10 Posts |
Hey Fat Freddy,
Thanks for pointing that out, I corrected the thread.
|
|
Moderator
 United States
188213 Posts |
 to the Community! Keep in mind you will not be able to send or receive e-mails via the forum until you are an established member. Feel free to share as many details as you are willing to in this thread. It is definitely an interesting idea. |
|
New Member
Belgium
10 Posts |
|
|
Pillar of the Community
United Kingdom
4208 Posts |
Thats a fantastic idea! Unfortunately, a pauper like me cant afford gold.
Do you reckon an algorithm is possible? Someone with a set of calipres could measure their coin accurately, input the numbers (diameter, thickness) and purity (hopefully known for the coin and what the gold is alloyed with) and let the algorithm work out the recquired tone. Check it via the microphone to get a yes/no on authenticity.
Or, even developing it further, work out the purity of a coin based on its tone (perhaps not possible due to different alloying). A bit more work would allow someone to get a set of purity ratings - content of gold when alloyed with different metals.
But I know nothing of this, perhaps its not possible to do such a thing. I'm sure the shape of the coins surface has some effect, no matter how small, on the tone.
|
|
Pillar of the Community
United States
807 Posts |
Certainly the speed of sound varies in various alloys, so it should be entirely possible to differentiate them by acoustic signature. The best test would probably use a transducer to generate a known acoustic signal, rather than dropping the coin (from a height which is not calibrated to deliver a specific amount of energy) onto a surface of unknown properties. This would also help to reduce the confusion generated by coins with lamination or inclusion errors, which can 'sound wrong' despite being of proper weight & composition.
|
|
New Member
Belgium
10 Posts |
@BenByfield: That would be possible. The problem is, however, that for small variations in alloys the material parameters that govern the acoustic fingerprint can vary. So the estimation would work best for alloys/materials that are well characterized, such as pure gold/silver.
What seems more logical, in my mind, would be having you create your own database of pings for whatever coin you have, and compare that with others. That way you could compare the composition of your coin with other people, if you both take the time to ping your coin. That's currently not a feature, but would that be useful?
|
|
New Member
Belgium
10 Posts |
@publius: The standard ping test works just fine. Just balancing the coin on your finger and tapping it slightly with a hard plastic pen for instance.
|
|
Moderator
United States
188213 Posts |
|
|
New Member
Belgium
10 Posts |
Wow, I'd love to get in touch with wert about that. Right now we're supporting three bullion coins. But this could definitely be interesting for all types of coins. What do you think?
|
|
Moderator
United States
188213 Posts |
Hopefully he will see this thread. I posted a link to this in the other one to get his attention (if he subscribed to it, he will get an e-mail).  |
|
New Member
Belgium
10 Posts |
Thanks :)
Thing is, I know almost nothing about numismatics. But the core principle on which the technology works should apply to all coins. I would love to hear what some real numismatists would deem useful.
|
|
New Member
United States
25 Posts |
I'd really be interested in hearing how you did this. Do you FFT the recording and search for harmonics at expected locations (within a minor threshold of frequency of course), and match that against some known profiles?
Do you feel comfortable sharing your code?
|
|
New Member
Belgium
10 Posts |
Hey NCSU,
Thanks for your interest. What you're describing was more or less our first approach. The problem, however, is that pinging a coin usually only excites 2 to 3 harmonics. Particularly in the case of 2 harmonics, there can be overlap between different coins. So although simple thresholds for the harmonics would give you a decent sensitivity (true positive rate), it would also involve a significant false positive rate.
For instance, purely on harmonics, the krugerrand and the maple leaf match up rather nicely. Which would require you to delve deeper. More specifically one needs to harness the information in amplitude, frequency and time domain.
I can answer your questions on specifics, but I'm sure you understand we'd like to keep our code, ours.
But in terms of questions, by all means, shoot away.
|
|
Valued Member
United States
337 Posts |
This begs the question, "How do the 'Cash for Coins' shops determine the composition of a coin?" Do they use a device that determines the gold/silver content? Would your idea be able to spot counterfeits?
|
| |
Replies: 19 / Views: 3,217 |
Posted 07/31/2013 08:39 am
