I have seen a project on the internet with an iPhone which was using a table as a keyboard : the iphone's accelerometer would detect the key touched via vibrations : http://goo.gl/5OpSC
Is it possible to make this project with an arduino and an accelerometer. Do you have any tutorial or articles?
It is a fake project only imaging what he wants. But technically it is possible to make such a device with I think 3 acceleration meters.
These sensors should be time coupled and by comparing the timing of when a pulse comes in one can determine the position of the "impact"
I wrote a similar sketch quite some time ago which may be helpful to get started.
It is discussed in this thread - Arduino Forum -
No it's not a fake this app does work, and some students of MIT managed to do great job : http://philipmarquardt.com/post/18048824898/publications
unfortunately it's a too theoretical for me, I'm rather searching for some pieces of code.
Excuses, you are right it is not fake, My French is not that well, I now used translate to read the article completely .
Also read the slides of Philip Marquardt.
To replicate this on an arduino you need at least a MEGA, possibly with extra memory to hold the database of words and their trained equivalent for the Neural network he uses.
(Part of ) the concept Philip presents is to detect Left and Right hits on the keyboard which generate a pattern which resembles the expected pattern for a given word in the dictionary. For this a neural network is used - not trivial for an arduino to do .
Think you need to experiment if an arduino can detect the patterns described by Philip. ONly then you can say if an arduino can do it this way.
Maybe I can use the arduino just to acquire the accelerometer's signal and then make the computer do all the work.
So you think I should work with pattern recognition?
I haven't seen anyone detecting key presses using impact detection, although I suppose it works for bullet hit detection so not impossible. I've seen somebody doing something similar by projecting a keyboard on the desk and then using an optical sensor to detect which key was being touched, and that seems likely to be easier to implement - but not an Arduino project.
That's how I understand the slides from Philip Marquardt. He trains a neural network to recognize the patterns coming from the sensors. It is a good idea to let a PC do the number crunching.
Do you know this one? - laser keyboard - YouTube - it is already a few years in the market.
Yeah I do know this one unfortunately it is a bit harder to build one