The Leap: Gesture Control Like Kinect, But Cheaper and Higher Resolution 192
MrSeb writes "It seems Minority Report-style computer interfaces might arrive a whole lot sooner than we expected: A new USB device, called The Leap, creates an 8-cubic-feet bubble of 'interaction space,' which detects your hand gestures down to an accuracy of 0.01 millimeters — about 200 times more accurate than 'existing touch-free products and technologies,' such as your smartphone's touchscreen or Microsoft Kinect. Unfortunately Leap Motion (the company behind the Leap) is being very tight-lipped about the technology being used, but it's probably some kind of infrared LIDAR (radar, but using light), or perhaps a high-resolution version of Kinect (which only uses a 640x480 camera). It's available to pre-order for $70 — and developers can register for a free device + SDK."
Fishy... (Score:5, Interesting)
I'm having trouble understanding exactly what kinds of technology this device is using to obtain accuracy on the level of 10 micrometers for $70. On the website they only state:
Leap Motion technology is a breakthrough in computer interaction, using a patented mathematical approach to 3D, touch-free motion sensing and motion control software that’s unlike anything that currently exists on the market or in academia. Developed over the past 4 years, Leap Motion moves far beyond the current technologies designed for distant arm waving.
But that say a whole lot of nothing... Why are they being so coy about the technology behind the device? According to cnet [cnet.com], the company says:
"It's not as if we're using lots of processing power or some new hardware that just came on to the market," he said. "This is really about a fundamental scientific breakthrough, many Eureka moments that (Holz) stumbled through over four or five years of research."
So they want me to believe they came up with some magic algorithm, and not some new hardware tech? Because as far as I'm aware, the limitations in most sensors is hardware based, not software.
Re:What's wrong with keyboards? (Score:5, Interesting)
stock pump (Score:3, Interesting)
Just want to mention, these guys are currently funding up and looking to put a shine on their company. With the secrecy all you're getting is their own spin, no actual info, and they are very much doing it do create buzz for their funding.
This is not a tech story, it's a stock pump in disguise. I'm not saying it's a pump and dump, but this is a pump for sure. It seems to be aligned with the Facebook IPO news to try to catch more ears.
I say this because I heard this news last week through my parents, who have an aggressive stockbroker who brings them endless "get rich quick" stock pumps. And for some reason they don't get rid of the guy even though he's pretty much all the worst things you can think of in a stockbroker. Must be a friend of the family.
Re:What's wrong with keyboards? (Score:5, Interesting)
Why do we need all these fucking gestures and shit? I guess most people still haven't figured out how to TYPE.
Typing is an unnatural interface that we've just grown accustomed to. After a while, we've become quite good at typing, but it's still an interface that one has to actually learn to use.
A more natural interface would be speech, but speech recognition is still far from the reliability of a keyboard.
On a parallel path, the mouse, while much (MUCH) more natural than a keyboard, can still get better. As soon as 2D displays are really replaced by 3D, mice would have to become 3D also.
The first problem was that 2D display (screen) and 2D gesture recognition (mouse) were easier than speech recognition. And now we are in a similar situation with 3D mice (this) and 3D screens (still crap).
The beauty of it all it that they'll all eventually converge in the same spot. Thought input and though output.
Is 2 mm accurate enough? (Score:4, Interesting)
0.01 mm * 200 woudl be 2 mm for Kinect. That seems accurate enough to me. It's pretty darn small.
Re:Fishy... (Score:5, Interesting)
I'm having trouble understanding exactly what kinds of technology this device is using to obtain accuracy on the level of 10 micrometers for $70. On the website they only state:
I'm having huge difficulty understanding how this is getting rolled out for video gaming instead of manufacturing.
A 3-d CAD "tasting" probe that goes in place of a cutting tool and touches what you're working on to measure its dimensions is about that accurate, very slow, requires some setup, and in manufacturing we pretty much don't care how much it costs (In a world of $100K milling machines and $30/hr CAM programmers, don't really care if its $70 one time cost or $7000)
If this isn't vaporware, how come I haven't heard about this tech destroying existing CAM monitoring/testing sensor systems?
Heck, 10 micrometers with low enough latency for gaming is enough to close the loop on a servo system.. imagine that, a CAM servo controller that doesn't need encoders. Weird but it could happen. Not to mention integrated OSHA detection of people entering the envelope or detection cataclysmic tool failure (snapped off).
I should be hearing about this making motor driver manufacturers and DRO manufacturers quake in their boots.
Re:What's wrong with keyboards? (Score:5, Interesting)
Speech is an unnatural interface, that one has to learn to use. It is also slow, inaccurate and cumbersome. Those are the problem when using it to interacts with other humans, trying to use it with computers is even worse.
The mouse is less natural, try to show an old person one. They will prefer the keyboard. 3D displays will not
Re:Developer link (Score:5, Interesting)
Considering they're not one, not two, but NINETEEN versions behind in their OpenSSL software (currently at 0.9.8x) AND they're running FrontPage extensions, I have little confidence in their online process for creating accounts and placing orders. Oh, and they're 2 versions behind on Apache as well.
Apache/2.2.19 (Unix) mod_ssl/2.2.19 OpenSSL/0.9.8e-fips-rhel5 mod_auth_passthrough/2.1 mod_bwlimited/1.4 FrontPage/5.0.2.2635 Server at www.leapmotion.com Port 80
Re:8 cubic feet... (Score:4, Interesting)
FOV is 57 horizontally and 43, with a maximum range of about 20 ft. I don't know how much volume that is, but it's significantly larger than 8 cubic feet.
Come on /., lets do some redneck engineering estimating with VLM down here at the high tech redneck (server) farm.
Lets wiggle your numbers to something smaller because I'm a lazy SOB and assume kinect only gives you a square of 40 degrees and only 1 foot deep precisely at 20 feet. Kin we all take a swig of grannie's recipe outta mason jar even tho its kinda early in the morning and agree that my grotesque simplification is a profound lower bound? Its going to be way huger than this estimate.
So 40 degrees up and down is really two rather acute right triangles of 20 degrees at the pointy end and, as you say, 20 feet on the adjacent side. Essentially we wanna solve for opposite side, times two because there's two right triangles, and square it to get square feet, and call it good because we're only looking at the one foot wide layer at 20 feet away (the true volume is way the heck larger, this is just a lower bound).
OK, sstill with me here? trig, um, 30 years ago, thats what "soh cah toa" tangent is opposite over adjacent. So tangent 20 deg = x/20. Tangent of 20 degrees is about a third... you can either trust me on that because I'm old, or you can GOOG it. Some basic algebra shows 1/3 = x/20 is the same as 1/3*20 = x in other words its a bit more than 6 feet, 6 * 3 being about 18 and 6*4 being too big for 20. So times 2 because there's two triangles means 12 feet. 12 feet up n down by 12 feet left n right at that 20 foot distance is 12*12=144 square feet. So we know it looks at the whole volume (does it?) so considering just that 1 foot shell as a minimum is 144 cubic feet.
Lets think backwards here to check. So if I imagine around a 40 degree triangle flying outta my eye, and look 20 feet away, can I see the ceiling? Yeah, I guess so. So I probably did the math correctly.
As a ridiculous upper bound, the upper bound of this pyramid must be smaller than a cube of 12*12*20 feet, right? So its less than 144*20 = 1440*2 = 2880 cubic feet.
So a kinect looks at between 144 and 2880 cubic feet in volume. This took an old engineering mind about 10 seconds to figure out and 5 minutes to type. In summary yeah its way the heck larger than 8 cuft probably 2 orders of magnitude bigger. But not 3 orders of magnitude bigger.