So we all know of all the different setups; we have a lot of them, in fact! For example, there’s DI, DSI FTIR… and the list goes on and on. But all of them are so bulky, am I right? For the webcam to view the surface or the whole surface I should say, sometimes takes a good 3 feet depending on size. Well, I didn’t like that idea. Bulk is not my style. Instead, I got to thinking about a new idea, and I came to call it MTXY. In theory, I believe it will work. Here’s how it is set up:

1.) acrylic, with IR LED’s or lasers along the sides, just like a normal FTIR setup
2.)LCD matrix with IR filter removed
3.)some type of IR stand still material (something to make a beam of visible IR light in itself to the camera) —webcams are on the sides
4.)LCD backlighting

The cameras would capture the sides of the acrylic and based on the horizontal rule of the emitted IR, it will put the X and Y coordinates there with the proper functionality. Of course the program deciphering would have to take just the top layer of the camera’s image so that it would process coordinates for just the width of the acrylic… (or would it matter ???)

I’ve included some Paint mockups as well. Maybe these will clarify what I’ve been thinking about over th apst few days. I just have one problem with this setup; What am I going to use for #3?? I can’t think of anything right now, but I’m sure maybe one of you all will think of something...(maybe). Hopefully this could be a working concept some day and make mutlitouch displays small enough width-wise to the point where they’re the size of a regular non-modified LCD monitor.

Please tell me what you think, I’m very interested for some feedback, even if it is negative… :D
Thanks for reading!

Hi hillbilly,

Very cool idea. A big negative aspect should be occlusion, if you have two ore more blobs in a vertical or horizontal line they cannot detected well.
Perhabs you can go a bit further and take lines of photodetectors at the sides, so the setup gets even more shrinked.

For the material… i think endlighten can do the job. You can order a sample sheet at http://www.plexiglas-shop.com for about 5€ (no shipping cost to germany).

I’m working on a test Multitouch setup for myself (14,1” TFT with Endlighten as backlight and IR illuminator), and though a lot about making the setup more smaller.

There is a possibility to use LEDs as photodetectors.

a similar approach is here:
http://nuigroup.com/forums/viewthread/3532/

I think your setup may has occlusion issues, have you think of that yet?

EDIT: oh, well, i think Knochi said it better

hi i was thinking about something similar time ago and made
some drawings and experimented a bit.
my attempt was with LEDs ontop the surface and the cams in the corners.
i was tracking the shadows.

biggest problem was occlusion, so this will do singletouch only.
i did not tinker further because i faced many other problems with
this so i focused my time now on my normal table.

the only good thing i found out was that it is not much affected by ambient IR.

i understand that occlusion problem, but: if you see a “blob” on the x axis and you don´t see one on the y-axis, you can assume that there are 2 fingers parallel to the y-axis, right? So maybe you can just assume where the finger is, to complete the values of the 2nd blob…
oh my english sucks you know what i mean?

edit: just added a picture for better understanding. as you can see, 3 fingers should be possible, or am I getting something very wrong?

You would have to use more cameras or have them in different locations. The occlusion happens in a cone shape past the object blocking the ir so that there’s more than one possible location the occluded object could be.

Yeah, but how to assume the position of the left blob?
It can be “behind” the upper or behind the lower blob.

A solution can be to lit the LEDs one after another and synchronize this with the camera. But due to the low speed of the camera you will get problems…
Perhabs somethings adaptive… if the controller detects hided blobs it turns of the LEDs which lights the right blobs and so he can “see” the left one.

but if you use photodiodes or something similar, the cone, if there is one, will have a very small angle…

@Knochi && Gustov : I can see both of your points of view…

It was a simple idea. I just hoped to improve the thickness of MT’s. Anyway, maybe I’ll try this one day if I ever get to disassembling my 23” monitor and getting some acrylic… and IR LED’s ... too… much.... to do!

Hillbilly,

Referring to your side image, you may have problems if the lcd is located between the acrylic and the “mystery layer”. The lcd basic blocks or allows light to pass through on a per pixel basis depending on the image being displayed. You would need to do some tests to see if the passage of IR light through the lcd is affected by the image on the lcd screen. If it is, then you could rearrange the order of your layers like so:

touch here
-----------------------
LEDs ---> | acrylic |
-----------------------
| Endlighten | >-- photosensors
-----------------------
| LCD |
-----------------------
| backlight |
-----------------------

With the layout above, you wouldn’t even need to disassemble the LCD screen.

Basically, the LEDs bathe the acrylic in IR light, when the user touches the acrylic the FTIR effect reflects the IR light from the user’s finger down through the acrylic to the Endlighten (your “mystery layer"). The endlighten causes the IR light to be reflected in all directions, including towards the edges of the endlighten where cameras placed on the top and side edges detect the change in IR light intensity, which is then filtered and analyzed to determine the X,Y coordinates of the touch location.

Other things to consider:

* This setup eliminates the MT Table height problem, but introduces an MT Table length and width problem.
* If cameras are placed at sufficient distance to see the entire edge of the table, how will you block out the rest of the camera’s FOV so it only detects the thin edge of the table and is not overwhelmed by “ambient” light.

Both of these issues might be able to be addressed by using a linear array of photosensors mounted directly to the edge of the endlighten in place of the cameras shown in your diagram. If you use a dense sensor array with each sensor shielded so it has a very narrow FOV, the occulsion problem is mostly eliminated as described by Vincent_vega.

Also, you might need some kind of coupling layer between the acrylic and the endlighten to optimize the light transmission between the two layers.

Pretty cool if it works. Sort of a side-scanning FTIR.  I would image it would take a lot of experimentation to get the right combination of materials and the right setup.

Oops,

Here’s a do-over of the diagram:

~Touch Here~
-------------------
| ~~acrylic~~ | <-- LEDs
-------------------
| ~Endlighten | >-- photosensors
-------------------
| ~~~LCD~~~ |
-------------------
| ~backlight~ |
-------------------

I’m new, so maybe I’m over-looking something, but isn’t the problem with occlusion not that you can’t see where you’re touching, but rather that you get false positives? So if you had three contact points, like in the example above, the computer would assume a fourth one… I altered one of those pictures to show what I mean.  And actually, if you just had the upper left and lower right contact points the computer would not be able to distinguish that from having the four points.

But what if you had x- axis, y-axis and a diagonal-axis as well? That would solve the majority of problems for occlusion, although processing the inputs might be a bit more hairy. If you used two cameras in the top corners and one in the top center with a 180 degree lens you could even keep them all really close to the edges of the screen.

The same principle could be used with photodiodes, you would have to dedicate two sides to the diodes for the diagonal, but I don’t see why this wouldn’t work. (except that you would have to build it from the ground up, program the input boards and get it to interface with your computer, which would be a great deal more work than working with cameras interface with your computer automagically)