Linked by Thom Holwerda on Tue 10th Jul 2012 01:24 UTC
Microsoft "Microsoft and Perceptive Pixel Inc. (PPI) today announced that they have entered into a definitive agreement under which Microsoft will acquire PPI, a recognized leader in research, development and production of large-scale, multi-touch display solutions." Yes, Jeff Han is now a Microsoft employee. This demo still amazes me - from 2006. Before the iPhone. Before Android. Before the iPad. Remember that the next time you wind up in a discussion about who supposedly invented what.
Thread beginning with comment 526428
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[6]: I built one too!
by Alfman on Wed 11th Jul 2012 21:14 UTC in reply to "RE[5]: I built one too!"
Member since:


Do you think acoustic reflection is that big a deal for software to handle in your model?

Granted, I don't know how much attenuation to expect, but I'd be surprised if it's difficult to differentiate between the primary and reflected signals. My guess is that secondary waves would loose high frequency components far more quickly than low frequency components.

Instead of trying to absorb/eliminate reflections, we may be able to use the information from reflections to our advantage. Consider a sensor on the right hand side of the table that picks up a pulse, shortly followed by one on the left hand side. Ordinarily the algorithm would measure the time lapse difference and register a corresponding touch on the right side. However the pulses may have been triggered by environmental sources off the table.

If the reflection isn't too week, we might be able to prove that the pulse originated on the table because from anywhere else the timing of reflections would be off.

Instead of attempting to write an explicit algorithm for this, it might be easier to use a generic pattern matching algorithm and perform a hough transform across it. This would produce a 2d graphic with intensities corresponding to the areas that have the strongest matches. This could be projected onto the table so that it gets lit up in the spots where it is touched. Then try to fool it with external noise.

Edited 2012-07-11 21:19 UTC

Reply Parent Score: 2

RE[7]: I built one too!
by Neolander on Fri 13th Jul 2012 10:47 in reply to "RE[6]: I built one too!"
Neolander Member since:

Well, theoretically-speaking, a few reflections on the sides of the screen from time to time are not that big of a deal. Without even going into waveform analysis, they would be detected as taps coming from outside of the screen, which can be eliminated by means of simple bound checking.

The first potential issue with reflections, however, is when you have lots of reflections going on with little damping. In this case, standing waves of significant amplitude would form inside of the screen after a short while, and potentially bring a significant amount of noise into the mics for a few seconds (think of a guitar string), which is bad news.

Another potential issue is reflection on the back and the top of the screen. These should probably be quickly damped, but if they aren't, they would result in the detection of multiple taps on a direction of space which the mics aren't sensitive to. I don't know if there would be an easy way to ignore those, maybe an heuristic like "Ignore taps which originate from nearly the same position within a delay smaller than 1 ms" could work.

I don't have enough of a signal processing and acoustics background to know whether your proposal to discriminate taps from the outside from sound from the inside would work, but spontaneously I would think that once an exterior pulse has hit the screen at some point, hard enough for mics to detect a signal, the perturbation would propagate inside of the screen just like a regular tap.

Reply Parent Score: 1