I was wondering if anyone else had come up with this idea yet. Or if anyone had already tried this and found it didn’t work. This is not a Cam and IR setup. In addition, it would lend itself well to either hardware or software blob detection. At least better than cam and IR does. See attachment for a picture as to the setup.
A plane of light would still be produced just like in LLP or a typical FTIR setup could potentially work. If fact, depending on your FTIR setup it may only be slightly difficult to perform this upgrade, once the extra materials to do this are available. Then along the edges either LEDs with a negative voltage and current, or photodiodes could be placed to receive the IR light. Whenever a finger is placed in the path of the light the voltage/current produced by the photoreceiver would be diminished. These photoreceivers could then be fed to an ADC and then processed. As an output the point locations could be transmitted.
The below sketchup assumes a 5mm led is used with a 2mm spacing between each LED. The 2mm spacing is probable a little optimistic.
The circles placed along the edges are the LEDs.
The boxes at the corner are the lasers(these are not to scale!)
The inner most box represents the touch surface.
The larger outer box is the path the lights from the ideal lasers.
everything other than the fingers and lasers are to scale.
Well i was originally thnking along the same lines as this but then i realized you would have the problem of occlusion in some cases… if you figure out a way to perform this with out occlusion then this technique would be awesome…
This way have the same problems, as LLP have. No fiducials in simple way and, as Taha said, occlusions. I’ve made little model of this. Previously I think you need specially placed fingers to get this problem, but it can’t even handle simple 5 finger touch with one hand.
There is a picture of this problem.
(little point on edges are shadows from fingers, big ones - centers of shadows, bold lines connects lasers with centers of their shadows, little black dots are retrieved finger positions)
In this example it is possible to find Grey, Yellow, bad Red and bad blue. Green finger is hidden. Also in this example you can not determine sizes of fingers.
First, in what real application are you ever going to have 5 fingers in that close of proximity to each other? I can’t imagine ever needing 5 fingers that close together.
Second, not being able to do good fiducials isn’t a requirement for everyone. I think reducing the size of a table is more important than having fiducials.
Yes, occlusion is a potential problem, but the requirement is to have extremely close together fingers, which in reality wouldn’t be that common. There is always the chance of it occuring, but what is the likelyhood? Does anyone have an educated guess?
I’d like idea of thin surface, too.
And of cause, fiducials is not necessary for everyone. But this is good feature.
It is not necessary to put fingers together. But fingers are close to each other, so they can be put in such position
Also You can achieve this using other positions. Yes, with less than 5 fingers it is not possible, and if you don’t need more, than it’s definitely good method.
I’d like to draw using four fingers - that’s fun.
In my case I additionally differentiate shadows made by different lasers. That’s simplify searching for fingers a lot.
You would alternate lasers that strike the same surface? I.e. topright and bottomleft can shoot at the same time since they don’t hit the same sensors?
Yes, lasers in opposite corners can pulse at the same time. In my example this done just for better view.
Did you try to make an algorithm to restore fingers by shadows?
I’d like to bring this back up. As the link shows the LEDS would be set behind the screen, so I don’t see the problem of occlusion. There aren’t any lasers involved at all. The LEDS are wired to emit light and sense light. Watch that video, looks pretty cool. Would it be possible to use these LEDs as a backlight for an LCD screen while also using them to detect multi touch events? This method seems nice because it eliminates the need for a projector and a webcam, which would make it possible to create some really thin screens.
I do not understand fully. Why would you need lasers, as it appears in the video there are no lasers. Do you have any schematics on how to wire the LEDs? If it could happen, I like EFNX’s idea to light your LCD with the same LEDs that are sensing your blobs.
The idea, described in the first post, is not the same as on the video. But they both are thin and don’t need camera. And this is the only common between this ideas - methods of finger finding are different.
okay, it makes sense now. Can anyone give any more info about the video or is anyone doing any investigating on that method? I would imagine the main downside is the grid/size of LEDs might make drawing a smooth line impossible unless you had small LEDs or a pretty large screen.
Actually this will work in my eyes, I was thinking of a similar setup except instead of photodiodes I was thinking about a camera strip (they don’t exist yet in my knowledge). But that was before I knew about photodiodes so now I will change my idea to include photodiodes instead of cameras. I created a thread here: http://nuigroup.com/forums/viewthread/3359/
I don’t think occulsion can occur if setup correctly with the software, look at my answers in my thread. Occlusion will occur with multiple people, but with 1 person it is almost impossible to have occlusion with the software setup correctly. EVEN if you put all 5 fingers right next to each other! Because LOGICALLY you can figure out the points which is what the software automatically would do! And it is easy to do because there is only 1 possibility (if you have normal hands) so its not a “guess” game. But for this to occur the data for EACH side must be separated and made into like a double X-Y grid. Look at my thread:
What about resolution??? I should think this would be a very jerky and not fluid at all.
There are plenty of potential gestures that would cause occlusions.
Sure… you can sell yourself into the idea that it will all work, but I think it’s better to acknowlege the potential problems
and solve them during the design phase instead of leaving it as a limitation to just be worked around.
The disadvantage of both these techniques [lasers with a perimeter of diodes or a 2d array of diodes] is the resolution. The advantage would be speed and a thin form factor. The diodes you use for backlighting the LCD screen could also act as the sensors. But the best resolution possible is going to be given by using the smallest LEDs, which can be made very, very small, but at that point it isn’t practical for our application due to the incredible amount of work and difficulty in assembling a 2d array of 1000+ LEDs.
I think a better way to get thinner screens is to use a smaller number of IR LEDs in an even spaced 2d array, along with a small array of low-resolution cameras. Where one acquires cameras of this sort, I don’t know.
