Hi, I am currently trying to build a multi-touch music application. I have read complaints about many multi-touch methods (with touchlib as the software) having very high latency/slow response time. This is not good for realtime music, as I would like to have things equivalent to a realtime onscreen piano and a drumpad w/ drumsticks. My first question is are there any methods out there already that do respond very fast at both the software and the hardware end? Also, if the first question is a no, then would attaching IR LEDs on gloves, drumstick, etc., improve speed? I was thinking it would look something like:

| <-- Finger (in glove)
o <-- IR LED at end of glove finger
-- <-- Diffuser (So the IR LED’s light is only shown when it touches the surface)
-- <-- Acrylic/Glass

{} <- Camera, Projector, etc.

I was thinking that this method would produce better blobs which would make touchlib faster. I’m not 100% sure though. I’m more interested in if there any methods out there already that do respond very fast at both the software and the hardware end. Thanks!

Cool hardware idea, it won’t make things any faster though.
A quad core CPU and maximum RAM will help the most.

I have found that older CPUs coupled with small amounts of RAM don’t run TouchLib very well.

I tried a setup on a Sony laptop with an AMD Athalon CPU @ 1.53 Ghz with a wimpy 256 MB of RAM.
It REALLY lagged mostly due to the minimal amount of RAM. The amount of memory is definately a factor.

I have a Mac Mini (Core Duo) with 2gb of RAM that I can dedicate to just TouchLib. Would this provide the kind of latency I require? (ie. A drumbeat with a drumstick on the surface must make a beat/sound almost instantly) Thanks!

I thought initially when I became interested in MT that a glove or something like that , like your saying would make sense seeing how at the time all the problematic issues were generally the same with MT. But a key goal I think is to make MT as easy to use as possible not only for one user at a time but multiple users if nesessary. How many gloves would be available or who wants to wear other peoples gloves , would be the issue, maybe....
I like that you though of that, it’s a very direct problem---solution approach . I think theres room for disscusion on your idea.

zachthehack, the latency doesn’t come from the MT technology. As Tinkerman said, using gloves or anything similar won’t improve speed at all. MT technology (DI, FTIR, etc) are all low latency techniques. The latency occurs starting mainly with the camera. If you want low latency, then a firewire camera would be better to start with. Then, you need something capable of high frequency (framerate). A 30fps would probably not be good enough for the speed you want and at least 60fps would be needed. Then in addition, the latency/processing speed of your tracking program (touchlib, reactivision, etc) has to be considered, also the communication latency between touchlib/other and your application, and then lastly your soundcard (which can also add large latency). With drumming, very low latency is required. You’re not going to be get anything low enough in latency for a while probably. There’s no way you’ll be doing drum rolls or swiss army triplets with a MT screen anytime soon. You’re asking for something that even some MIDI keyboards have a hard time with (very low latency).

Though, I wonder… One thing that always struck me about multitouch is the fact that it requires such an overhaul in user interface.  I was thinking about fashioning a glove that has triggers on the fingertip.  When slightly depressed, they give off a small amount blob and as the pressure increases it grows to the size of a whole normal finger.  The amount of travel would be slight, say half a centimeter.  Now, if someone pulls up a multitouch keyboard on their screen, they can type on it without having zero tactile feedback, and still use their hands naturally on the screen.  It wouldn’t allow for using your whole hand, but I think that’s rarely the case in multitouch.  Granted, I’ve only recently looked into any of this, so perhaps I’m not informed enough to know that this is incredibly difficult or even unnecessary.  This is my first post btw

Maybe a part of the solution to keep “usual” multitouch in the process is to link the software process entirely. Having for example predefined zones (selected by an application), this way touchlib don’t have to scan the entire screen so this computing part is really fast (I don’t know maybe this kind of feedback between touchlib and an application is already made ?). Of course there is still the high-end look of applications which isn’t the common case with usual music softwares.

Of course improving the hardware (60 and more FPS etc...) is a real bonus but there is a price.