Data/Algos for most realistic orientation/position tracking

geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Data/Algos for most realistic orientation/position track

Post by geekmaster »

Although Project Holodeck currently uses optical position tracking, a Hydra-like magnetic absolute position sensor would be great for the consumer Rift if Oculus can extend the range and work out some of the kinks when using a Hydra for head tracking, as mentioned by John Carmack.

Because Oculus is selling a product that is not a DIY kit, there are patent and licensing issues to deal with if they want a frequency-based magnetic tracking solution (as licensed to and used by the Hydra). They may have to work their way around the IP, or use the expired-patent DC magnetic tracking method.

There has been some information leakage about what they may be working on for absolute head tracking. There was this tweet:
OculusVR has snatched Nirav Patel from Apple & will use Adjacent Reality tech to track head position.
... which may have something to do with this thread:
The Adjacent Reality open source wireless head/hand Tracker
http://www.mtbs3d.com/phpbb/viewtopic.php?f=138&t=15184
Knowing that Nirav Patel is nrp, we can put those two bits of information together and make some educated guesses... :D
Pyry
Two Eyed Hopeful
Posts: 85
Joined: Mon Aug 13, 2012 5:55 pm

Re: Data/Algos for most realistic orientation/position track

Post by Pyry »

I don't remember where I saw it, but hasn't that adjacent reality IMU connection been known for months? In anycase, if Oculus wants magnetic position sensing, they would probably be best served by working out out a deal with Sixense rather than trying to do it themselves while avoiding stepping on patents. The Hydra isn't exactly selling like hotcakes, so I would think Sixense would be pretty reasonable about a licensing deal.
User avatar
daniel2e
One Eyed Hopeful
Posts: 17
Joined: Fri Feb 15, 2013 5:06 pm

Re: Data/Algos for most realistic orientation/position track

Post by daniel2e »

geekmaster wrote: Knowing that Nirav Patel is nrp, we can put those two bits of information together and make some educated guesses... :D
It sounds like there is interesting news about on horizon regarding this integration, though since Oculus is a proper business now I'm not sure we'll find out until the consumer version is hitting the assembly line. Not that I blame them, Oculus is more or less inventing this market and absolutely deserve the inside track, so if they need to play things close to the vest more power to them.

That said, if they are having trouble on the software side, I'll have my fingers crossed for a dev kit add-on and some hooks in the SDK ;)
PatimPatam wrote:Also as daniel2e suggested, if it can serve some other purpose like as an information gathering device, i'll be happy to help!
When the kits roll out I may hold you to that :)

I'll bet you can get very accurate tracking from all those LEDs, better (or at least, more straightforwardly) than any magnetic field solution, particularly if you don't have to worry about latency (and you wouldn't). If so, that means you can generate a data set that allows objective evaluation of different fusion algorithms and sensing platforms. That would be a very valuable tool, and at the moment you may be only person with that capability.
User avatar
FingerFlinger
Sharp Eyed Eagle!
Posts: 429
Joined: Tue Feb 21, 2012 11:57 pm
Location: Irvine, CA

Re: Data/Algos for most realistic orientation/position track

Post by FingerFlinger »

I would expect Oculus to be using an "outside-in" optical tracking solution for position tracking initially. The magnetic trackers seem tied up in patent issues, and optical flow solutions will not be robust enough for a commercial release for quite some time. And I still think purely inertial solutions are totally out; people have been working on that problem for ages.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Data/Algos for most realistic orientation/position track

Post by PasticheDonkey »

simplest head position track solution i can think of would be a physical bridge between the head and a chair. basically three filament wires that you can assess the length of and judge angle from comparison. they'd be on reels so amount of reel turn would be the data set. if you add chair position for a swivel chair on top of that then it's a good sit down system. might need a way to determine angles of filament to avoid multiple solutions to same data.
User avatar
FingerFlinger
Sharp Eyed Eagle!
Posts: 429
Joined: Tue Feb 21, 2012 11:57 pm
Location: Irvine, CA

Re: Data/Algos for most realistic orientation/position track

Post by FingerFlinger »

Ha, that sounds very similar to a scheme somebody came up with for measuring locomotion a while ago. But I can't see it going onto a commercial product, and as you said, there are ambiguity issues. It would be neat to try though, maybe for a desktop handtracker instead of a headtracker. Force feedback anyone?
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Data/Algos for most realistic orientation/position track

Post by geekmaster »

PasticheDonkey wrote:simplest head position track solution i can think of would be a physical bridge between the head and a chair. basically three filament wires that you can assess the length of and judge angle from comparison. they'd be on reels so amount of reel turn would be the data set. if you add chair position for a swivel chair on top of that then it's a good sit down system.
I think using one spooled filament instead of three would be less problematic. The Mad Catz GameTrak used one filament spool, threaded through a joystick handle. I remember reading about somebody using a ceiling-mounted GameTrak to track head position for a DIY HMD and one hand.

Image

In that photo, the filaments extending from the joysticks can be moved in 3 DoF, tracked by the joysticks and spool rotations.

The GameTrak has been discussed at MTBS3D before:
http://www.mtbs3d.com/phpBB/viewtopic.php?f=138&t=15717
http://www.mtbs3d.com/phpBB/viewtopic.p ... 900#p77900

EDIT: I got a GameTrak from eBay for $0.99 (but it was missing the foot switch).
Last edited by geekmaster on Wed Feb 20, 2013 10:00 am, edited 2 times in total.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Data/Algos for most realistic orientation/position track

Post by PasticheDonkey »

yeah it's that kind of system that inspired me i just thought of attaching it to a chair behind the head. and couldn't remember it's name. i wanted three for absolute angle as well as position
Last edited by PasticheDonkey on Wed Feb 20, 2013 10:14 am, edited 2 times in total.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Data/Algos for most realistic orientation/position track

Post by geekmaster »

This guy used a head-mounted graphics tablet with chair-mounted stylus. It only tracks 2D roration though, and the Rift already has that:
[youtube]http://www.youtube.com/watch?v=hlLRq_biPOM[/youtube]

Cool hack though...

For cheap head tracking, you can use FreeTrack with any old webcam:
http://en.wikipedia.org/wiki/FreeTrack
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Data/Algos for most realistic orientation/position track

Post by PasticheDonkey »

i doubt i'm going to suspend a web cam in front of my swivel chair.
geekmaster
Petrif-Eyed
Posts: 2708
Joined: Sat Sep 01, 2012 10:47 pm

Re: Data/Algos for most realistic orientation/position track

Post by geekmaster »

PasticheDonkey wrote:i doubt i'm going to suspend a web cam in front of my swivel chair.
No, you have it "backwards" from what I was thinking. You could mount it just like your "filament" suggestion, but unlike having filaments attached to your head, tracking light rays would be less problematic. You could face the LEDs to the rear and put the camera behind the chair, attached to the head-rest. You would have to rotate your frame of reference though...
Last edited by geekmaster on Wed Feb 20, 2013 10:34 am, edited 1 time in total.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Data/Algos for most realistic orientation/position track

Post by PasticheDonkey »

ah yes that would be good. nay perfect for that operating scenario. it'd just be finding a clamp that would work for the variety of head rests that'd be the issue.
User avatar
daniel2e
One Eyed Hopeful
Posts: 17
Joined: Fri Feb 15, 2013 5:06 pm

Re: Data/Algos for most realistic orientation/position track

Post by daniel2e »

The Oculus twitter account just mentioned updates to their job listings, which brought my attention to this one.and this one, with the following excerpts:
We’re also researching positional tracking (optical and otherwise)
- Add new stereo vision analysis for motion tracking (hands, objects) to the Oculus SDK.
These seem to indicate they are learning towards optical solutions for position tracking, and although the head is not specifically mentioned in either quote I'd have to believe that is the priority. I wonder how, or if, this interacts with the revelation yesterday that every PS4 will be equipped with a stereo camera?
Pyry
Two Eyed Hopeful
Posts: 85
Joined: Mon Aug 13, 2012 5:55 pm

Re: Data/Algos for most realistic orientation/position track

Post by Pyry »

daniel2e wrote: These seem to indicate they are learning towards optical solutions for position tracking, and although the head is not specifically mentioned in either quote I'd have to believe that is the priority. I wonder how, or if, this interacts with the revelation yesterday that every PS4 will be equipped with a stereo camera?
I hope by "optical" they mean using a depth camera or fiducials (LEDs, AR tags) and not stereo. There's a reason all the consumer-available systems (Kinect, Asus Xtion, Leap, Intel & Creative's teamup the "Interactive Gesture Camera") use structured light depth estimation rather than trying to do stereo.
MSat
Golden Eyed Wiseman! (or woman!)
Posts: 1329
Joined: Fri Jun 08, 2012 8:18 pm

Re: Data/Algos for most realistic orientation/position track

Post by MSat »

Pyry wrote:
daniel2e wrote: These seem to indicate they are learning towards optical solutions for position tracking, and although the head is not specifically mentioned in either quote I'd have to believe that is the priority. I wonder how, or if, this interacts with the revelation yesterday that every PS4 will be equipped with a stereo camera?
I hope by "optical" they mean using a depth camera or fiducials (LEDs, AR tags) and not stereo. There's a reason all the consumer-available systems (Kinect, Asus Xtion, Leap, Intel & Creative's teamup the "Interactive Gesture Camera") use structured light depth estimation rather than trying to do stereo.
Kinect and the Xtion work on a different principal than the Leap (which does use stereo cameras), and different still from the Creative device. Not to mention that the new PS Move will also use stereo cameras.
Pyry
Two Eyed Hopeful
Posts: 85
Joined: Mon Aug 13, 2012 5:55 pm

Re: Data/Algos for most realistic orientation/position track

Post by Pyry »

Nobody seems to quite know how the leap works, but it definitely isn't just stereo matching on a normal image pair (this article confirms that it is not stereo, and that only one camera is necessary). I'm assuming it's using some form of controlled illumination. Likewise, the PS4 does have two cameras, but the controller has a glowing fiducial on the front. The Creative Interactive Gesture Camera (what an unwieldy name) is time of flight, you got me. The point is, you should use any of the many available active depth-sensing technologies, or just use fiducials, and not try to do stereo matching from normal cameras.
MSat
Golden Eyed Wiseman! (or woman!)
Posts: 1329
Joined: Fri Jun 08, 2012 8:18 pm

Re: Data/Algos for most realistic orientation/position track

Post by MSat »

Ok, I missed your point. But I will say that if they figure out a workable and robust system that extracts relevant information from a stereo pair, then why not do it? Sure, it wouldn't be easy, but who am I to argue if they get it to work?
badposture
One Eyed Hopeful
Posts: 3
Joined: Sun Feb 17, 2013 12:41 am

Re: Data/Algos for most realistic orientation/position track

Post by badposture »

Pyry wrote:There's a reason all the consumer-available systems (Kinect, Asus Xtion, Leap, Intel & Creative's teamup the "Interactive Gesture Camera") use structured light depth estimation rather than trying to do stereo.
What's the reason?
User avatar
KBK
Terrif-eying the Ladies!
Posts: 910
Joined: Tue Jan 29, 2013 2:05 am

Re: Data/Algos for most realistic orientation/position track

Post by KBK »

MSat wrote:Welcome to the forums, and great first post!

While there may be some methods and algorithms to increase the quality of data from the current IMU, there is only so much that is possible due to inherent limitations. What these limitations are and where they derive from, I don't know exactly (at least without knowing the sensor specs and the process by which the data is acquired), but they seem to be in line with what numerous researchers have experienced using similar sensors. From the looks of it, it would seem that orientation data is quite reliable in terms of both response time and accuracy. All three sensors work well together towards these goals, and low-pass filtering of accelerometer and magnetometer data is good for providing low noise orientation info. However, when it comes to translations, you're pretty much stuck with using just the accelerometer - low pass filtering is only viable if accuracy and sampling rate is high enough (as it turns out with most sensors, resolution decreases with an increase in sampling rate), the only things you can use the gygro/mag for towards these ends is to verify that there was no change in orientation, which is of marginal use.

As is, the Rfit's IMU on its own is unsuitable for translational tracking. There needs to be additional sensing systems in order to make such functionality possible - for which there appears to be no shortage of.
I can drop the noise floor on these devices, I suspect. It should have the desired effects. (a multi-pronged attack or set of solutions that work together)
Intelligence... is not inherent - it is a point in understanding. Q: When does a fire become self sustaining?
User avatar
ftarnogol
Cross Eyed!
Posts: 143
Joined: Thu Oct 11, 2012 7:38 pm
Location: Argentina
Contact:

Re: Data/Algos for most realistic orientation/position track

Post by ftarnogol »

I just found this.

http://phys.org/news/2013-02-technology ... eroes.html

I wish they released the algorithms/soft...
Image
<a href="http://phobos.psychologicaltechnologies.com">Link text</a>
Pyry
Two Eyed Hopeful
Posts: 85
Joined: Mon Aug 13, 2012 5:55 pm

Re: Data/Algos for most realistic orientation/position track

Post by Pyry »

badposture wrote: What's the reason?
People have been working on stereo for decades; it's one of the most researched areas in computer vision. Despite that, it doesn't really work that well (or, to put it more positively, it is a hard, far-from-solved problem with many open questions). You have a choice between fast algorithms that produce pretty bad depth maps, or slow algorithms that produce ok-ish depth maps. In contrast, the many depth sensors you can now buy are both fast, and produce pretty good depth maps.
User avatar
android78
Certif-Eyable!
Posts: 990
Joined: Sat Dec 22, 2007 3:38 am

Re: Data/Algos for most realistic orientation/position track

Post by android78 »

ftarnogol wrote:I just found this.

http://phys.org/news/2013-02-technology ... eroes.html

I wish they released the algorithms/soft...
That's pretty cool, but I'd like to see a video where they've mapped the bones to a 3D model so that you can see how accurate that is as they do a fly round (from different angles).
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: Data/Algos for most realistic orientation/position track

Post by cybereality »

That video looks amazing! Would love to see that tech used for VR.
Post Reply

Return to “Oculus VR”