Linux Kernel Module For Nintendo Powerglove

antistatickid writes: "I've dusted off some schematics for a simple parallel interface to the nintendo powerglove (circa 1990), and have written a linux kernel module for the device since none of the old code works anymore. I'm hoping to generate some interest in homebrew vr: the gloves are cheap, and can be used for things like controlling midi synthesizers with the wave of your hand (a demo of which I've included on the project page)."

Comments

Wrong use!

[Scutter]

Screw synths! You should be playing Black & White!

Wireless Mod?

[TibbonZero]

I wonder if there is anyway to mod the Powerglove even more to make it wireless? (I am thinking of performance live)
Just hack open a wireless Nintendo controller? And use the insides of it? Use batteries to power the glove?
Electronics on the compenent level isn't my type of thing, but I feel that it's possible. Is it?

Re:Bad Old Days

[antibryce]

A friend of mine used powergloves as part of a research project in college, and he said if you stripped out the sensors and mounted them on a pair of thin leather gloves they're MUCH more responsive. I dunno how accurate that is, but it would make sense that a huge rubber/plastic glove would affect things a bit.

Sweet!

[GoRK]

Very cool! I built this hack in like 1993, and it was hard to find a power glove even then. Now, it must be rather impossible.

A tip for hax0rs: The power glove is very SMALL (even the large one). I completely dispensed with the original glove that came with it to make mine. I took the control pad off and put a simple belt clip on it. Next, i extended the hand part and the ultrasonic sounders away from the controller with some 15 conductor cable. Finally, I sewed the finger bend sensors onto the fingers of a golf glove that went on the right hand and had the fingertips cut out (the original power glove is a lefty device.) Anyway, the idea was to get rid of the bulky garbage of the powerglove in order to make a little dataglove that i could still type while wearing.

I still have it here. Heck, I still have the monitor with the velcro on it! I'm very excited to break it out again and fiddle with this.

~GoRK

Real DataGloves

[f00Dave]

Why hasn't anyone built a *REAL* dataglove for the masses yet? The PowerGlove is a lame-ass mockery of a real 3-space input device and is only good for use in simplistic games or other 'toy' applications!

Wow, that reads like a Flamebait. ;-)

What I'd really like to see is a cheap-in-volume 'glove pair' input device (say 100$ for the MS version and 30$ for the Logitech one, like mice or keyboards) that would stream the positions of the fingers and hands over a hot-pluggable USB connection. I have a bazillion applications for that sort of device, and even a good headstart on a way to produce one on my own for about 300$ per pair (and a whole lot of time I just don't have). I'm sure *someone* has already had similar thoughts....

For reference purposes, my (rather fluid) specifications are for a system that:
- spits out positions of the fingertips accurate to 1 cubic mm or so within a cubic meter in your 'work area' (ie: a volume sitting above a traditional keyboard's location at a desk)
- tethered or wireless, as the case may be (wireless is an extra cost, of course, but not THAT much extra - it's mainly the short battery life that sucks for this)
- 60 Hz or better refresh rate for each of the sensed positions
- serial or USB input stream, similar to a 2D mouse's, only with a LOT more coordinates ... this is emminently compressible data, too, should bandwidth prove an issue (though there's always FireWire and USB2.0, I guess)

So, why should everyone have one of these? Well, I can't give away ALL my secrets, but people laughed at the mouse, didn't they? =) A 3D desktop metaphor requires a 3D interface device, and 'air mice' sort of suck. Wands are only good for limited applications ... remember light pens? (They're the same thing as a mouse, in a 2D sense, and you don't see many light pens kicking around today, do you? =] )

How would you like to type on a virtual keyboard, configured any way you want it to be, anywhere in space you chose to place it? How about a 20-DoF controller for videogames? Music synthesis with 20+ DoFs, each affecting a different component of the sound (left hand for timber and right for pitch, volume and sequencing)? Just as the mouse hardware drove the creation of a billion 2D applications, so will 3D 'glove' hardware drive a billion more.

But only, ONLY if it's CHEAP. If anyone knows an electrical engineer that wants to work on the hardware end of a project with me (I've got the hardware feasability, sample applications and reconstruction algorithms mostly worked out ... but no time to spend on implementation), give 'em my email.... f00Dave@bigfoot.com

God, that was a lot longer that I'd expected it to be. Must be the heat. =)

Re:Real DataGloves

[rusty0101]

- spits out positions of the fingertips accurate to 1 cubic mm or so within a cubic meter in your 'work area' (ie: a volume sitting above a traditional keyboard's location at a desk)
- tethered or wireless, as the case may be (wireless is an extra cost, of course, but not THAT much extra - it's mainly the short battery life that sucks for this)
- 60 Hz or better refresh rate for each of the sensed positions
- serial or USB input stream, similar to a 2D mouse's, only with a LOT more coordinates ... this is emminently compressible data, too, should bandwidth prove an issue (though there's always FireWire and USB2.0, I guess)


Not to dampen your intentions, I think they are admirable, however I do have a couple of notes on this for you. A meter is a bit arbitrary, and will give you problems with limits and data bandwidth.

For example, most people work en an environment where reaching a meter above their keyboard is only done when they are about to impart excessive forces on the keyboard in frustration. Unless you are doing something that requires position sensitive gogles as well, you are probably going to be better off working with half a meter vertically.

On the other hand reaching out to each side is not particularly unusual, and will easily exceed one meter side to side for most people. A range of either a meter and a half, or two meters would be safer.

This resolution of 1 cubic mm is also going to be expensive. I would think that it would make much more sense to vector track the hands, perhaps with a palm sensor which would give rough estimates of direction and speed while moving, then provide more accurate positioning data once stopped relative to the earlier movement. A surgon using such a glove is going to consider one millimeter to be aufull sloppy if he has to make an incision. At the same time, when reaching out towards the ends of our reaches, we are less interested in that 1 mm sensitivity. With a little bit of thought, you could use this area as broader spectrum navigation. Similar to using edge detection to move from one virtual screen to another, if you cross the edge of the sensors range, your virtual working area changes. If you are doing distance surgury, reaching into some areas would activate instrument changes.

Also of note is that the fingertips are rarely more than 150 mm from the center of your palm. You could easily use different resolutions for different fingers as well. As an example, you could use a sensitivity of 1 mm for your thumb, and .1 mm resolution for your index finger. You could also reduce the number of sensors required by recognizing that the ring finger is rarely as strong as the rest of the fingers, and eliminate that sensor. (for most movements, other than typing and musical instruments, the ring finger acts in concert with the pinkey.)

For purposes of the calculations of bandwidth I will use the dimensions you have provided however. You are free to use whatever of the ideas I have noted to finetune these. (given the fact that the fingers on each hand are always close together, you could compress the information by giving one finger's position at 10 bits x, 10 bits y, 10 bits z, then offset the remaining fingers from that position with 7 bits per dimension.

In any case, if you give each dimension a seprate holder, the smallest number of bits you can send per sample is 300. (10 bits per direction, [2^10=1024] three dimensions per finger, 10 fingers) multiply this by 60 samples per second, and you are running 18kbps. Even if you ad overhead, such as a stop bit every dimension, and a start bit for every sample, the bandwidth requirements are not high. At least not by modern communications standards anyway. The problem is that we do not have that many devices that are both moving (which will cause wire and fibres to ultimately breakdown) and sensing their environment that use this kind of bandwidth.

I suspect that whatever solution you put together will be regularly susceptible to failure due the the multiple moving parts required to track the hands of the user. You might be able to find a way to do it with fingertip and palm sensors that wirelessly communicate with each other, or that each communicate with a base station of some sort. One example would be a two camera system working with florescent fingertips that the user would wear. Similar in effect to a motion capture system.

Oh, well, best of luck to you in your venture...

-Rusty