Kinect technology thread

[patsu]

If you have time, you can google and find out. ^_^
I found the gamasutra.com article above using google too.

I think you should be able to find a depth camera video feed also.

EDIT:
Here's an outdated article that mentions 48 joints and dual 640x480 cameras:
http://electronics.howstuffworks.com/microsoft-kinect.htm/printable

It is possible that the GameStop page is also outdated.

 

[eastmen]

If you have time, you can google and find out. ^_^
I found the gamasutra.com article above using google too.

I think you should be able to find a depth camera video feed also.

EDIT:
Here's an outdated article that mentions 48 joints and dual 640x480 cameras:
http://electronics.howstuffworks.com/microsoft-kinect.htm/printable

It is possible that the GameStop page is also outdated.

I'm not so sure that this is the out dated info because one of the articles sources is from july 5th clearly later than the june 30th posting at play.com


amazon lists 640x480 cameras also
http://www.amazon.com/gp/product/pr...f=dp_proddesc_0?ie=UTF8&n=468642&s=videogames

 

[Shifty Geezer]

So we are agreeing there are only two cameras correct ?

However you say one is 320x240 and one is 640x480. However gamestop says
•Depth Camera 640x480 pixel resolution @30FPS

Check back through this thread. The depth camera sensor is 640x480, but the output resolution is a quarter of that because the PrimeSense tech uses quarter resolution (probably some degree of filtering). So both are correct, but the important figure is the 3D depth data used for skeleton tracking is 320x240. Kinect already has issues separating close players, so a later four player upgrade is very unlikely, as you'd need...6 metres (?) between you and your TV to fit everyone in view, and there'd be not many pixels from which to identify which bits are which.

 

[Arwin]

Remember that the depth perception works by an infrared projector projecting a grid onto the scene. The depth camera basically looks at the scene with a 640x480, and then processes the image to read how the grid reflects back to it to process the 16bit 320x240 depth-map. The confusion is understandable as the CMOS is 640x480, but the datastream is 320x240. So it's all correct, in a way.

 

[Panajev2001a]

I wonder how the frame rate affects things... PS Eye captures 640x480 feeds at 60 FPS... if you accept a 320x240 feed then you can push it at 120 FPS. Are we sure that Kinect's RGB camera is limited to 30 FPS when capturing 480p streams?

 

[Arwin]

I wonder how the frame rate affects things... PS Eye captures 640x480 feeds at 60 FPS... if you accept a 320x240 feed then you can push it at 120 FPS. Are we sure that Kinect's RGB camera is limited to 30 FPS when capturing 480p streams?

Affects what exactly? Both cameras on Kinect run at 30fps at 640x480. It's possible the RGB camera could do 320x240 at 60fps, but what's the use? And I'd expect the quality of the depth camera feed if it captured the projected infra-red grid at 320x240 to be almost useless - the grid would all but disappear and distant objects couldn't be seen. Perhaps if you wanted to do something close to the camera (though still at the min range for the projector which is something like 4 feet?) that required very fast movement and reflexes, then perhaps it could have a benefit.

 

[Shifty Geezer]

Note it's not a uniform grid, but basically AFAICT a random noise pattern. This is less prone to sampling errors than uniform patterns, according to PrimeSense's bumph.

 

[Arwin]

Makes sense. Anyway, being a pattern, it doesn't change the basic principles of needing to be able to see the pattern at a high enough resolution.

 

[Panajev2001a]

Affects what exactly? Both cameras on Kinect run at 30fps at 640x480. It's possible the RGB camera could do 320x240 at 60fps, but what's the use? And I'd expect the quality of the depth camera feed if it captured the projected infra-red grid at 320x240 to be almost useless - the grid would all but disappear and distant objects couldn't be seen. Perhaps if you wanted to do something close to the camera (though still at the min range for the projector which is something like 4 feet?) that required very fast movement and reflexes, then perhaps it could have a benefit.

Motion tracking in general. I was wondering if PSEye was actually overspecced as it can basically record at 2x the framerate given the same resolution (probably it is not as the fluid video streams should be able to help the motion sensors in Move).

 

[Shifty Geezer]

(probably it is not as the fluid video streams should be able to help the motion sensors in Move).

Definitely. We've talked about situations with Kinect like boxing where certain movements will be lost at 30fps. 120fps would provide excellent motion tracking, although lower performance in other areas like light and resolution. Given Move is an active lightsource, it could well be worth using the 120 hz mode in some action titles.

 

[bkilian]

Yeah, i just meant that giving more processing power (like pluging kinect on a Pc, or coming up with a new tracking system that's less resource intensive) could enable 4 player tracking even with 1st gen kinect

And again, where would you put them??

It would need a much wider viewing angle to support 4 players at any reasonable distance.

 

[scently]

And again, where would you put them??

It would need a much wider viewing angle to support 4 players at any reasonable distance.

Indeed.

 

[patsu]

Wii has 4 player games because some of their games don't need so much space. You can stay in one spot and play the entire game.

EDIT: 4P Wii Tennis
http://www.youtube.com/watch?v=U78ylX3hlvI

 

[AlphaWolf]

Wii has 4 player games because some of their games don't need so much space. You can stay in one spot and play the entire game.

EDIT: 4P Wii Tennis
http://www.youtube.com/watch?v=U78ylX3hlvI

They could also be sitting together on a sofa because the wii controller doesn't actually require anywhere near that much movement.

 

[Silent_Buddha]

Wii has 4 player games because some of their games don't need so much space. You can stay in one spot and play the entire game.

EDIT: 4P Wii Tennis
http://www.youtube.com/watch?v=U78ylX3hlvI

Even more important in the case of the Wii is that you aren't limited by the camera FOV. Since the camera is located in the Wiimote, you could in theory be at a fairly extreme angle to the TV and still be able to have the Wiimote see each of the IR emitters. Something which is a limitation of both Kinect and Move.

Regards,
SB

 

[patsu]

It depends on the game. I doubt this title uses pointer mode at all. The last 2 players will be handicapped. Like WiiSports Resort, they can implement this game based on sensor data alone. Since they don't need to run, absolute positioning is not required.

Similarly, 4P games for Kinect may be possible as long as they don't rely too much on the skeleton tracker. It is possible to roll you own custom tracker right ?

 

[Shifty Geezer]

Similarly, 4P games for Kinect may be possible as long as they don't rely too much on the skeleton tracker. It is possible to roll you own custom tracker right ?

I think we're getting into a semantics argument here. What constitutes a 'four player game'. Some are clearly thinking of Kinectg Adventures or it's ilk with four players, but of course four people sitting on a sofa and having to jump up could be tracked with the camera and would make a four player game.

The real point raised was the skeleton tracking tech. The PrimeSense comment said there's no intrinsic hardware limit to number of players, and to qualifty that, we need to point out that there are physical limits and data resolution limits, such that 4 player side-by-side skeleton tracking is implausible. That devs can use the whole glut of Kinect options to incorporate four players into one game is a side issue regards what sorts fo games could be made.

 

[goonergaz]

Even more important in the case of the Wii is that you aren't limited by the camera FOV. Since the camera is located in the Wiimote, you could in theory be at a fairly extreme angle to the TV and still be able to have the Wiimote see each of the IR emitters. Something which is a limitation of both Kinect and Move.

Regards,
SB

surely it has reversed limitations? doesn't the sensor on the TV has limited FoV but also the wiimote 'camera' too so there's still a fair chance the 2 won't see each other (in some games - e.g. when I played golf the club would read totally wrong) whereas with Move the ball means there's much less chance of the signal not getting the the camera, likewise you just need to be in view of Kinect for it to work?

 

[Nesh]

surely it has reversed limitations? doesn't the sensor on the TV has limited FoV but also the wiimote 'camera' too so there's still a fair chance the 2 won't see each other (in some games - e.g. when I played golf the club would read totally wrong) whereas with Move the ball means there's much less chance of the signal not getting the the camera, likewise you just need to be in view of Kinect for it to work?

I dont think it matters with Wii I think because it doesnt necessarily need to know where you are standing. Each controller is assigned to a specific avatar, so no matter where u stand AFAIK it will track the controllers movement to that virtual character. Only where the controller points in relation to the IR sensors is important for each player. Similarly when you play 2 player in Sports Champions' table tennis it wont matter if the guy controlling the right character is on the left and vise versa because the different color tells the system which character he should be controlling. From there on only the movement matters and if the led light is visible by the camera

 

[Silent_Buddha]

surely it has reversed limitations? doesn't the sensor on the TV has limited FoV but also the wiimote 'camera' too so there's still a fair chance the 2 won't see each other (in some games - e.g. when I played golf the club would read totally wrong) whereas with Move the ball means there's much less chance of the signal not getting the the camera, likewise you just need to be in view of Kinect for it to work?

Sure the Wiimote camera has a limited FOV. But it is controlled and pointed by the person holding it. Thus it's a non-issue for any player to be able to point it at the IR emitters regardless of where the player is. Well, except if something physically blocks line of sight to an emitter.

At that point it becomes a software issue of how robust the triangulation algorithm is with regards to larger angles of incidence as well as distance. Players on extreme corners would be much farther away from one of the emitters.

But the basic point is this. With both Move and Kinect, you're limited to a very fixed FOV within which all players (or move controllers) must reside in order to have any sort of absolute positioning. Move has a bit more leeway since it can rely on the inertial trackers within to have some degree of tracking.

With Kinect and it's absolutely reliance on the camera, it then becomes a limitation of having to make sure everything that needs to be tracked remains within the camera's FOV. Thus, the 2 player limit for all the launch titles is a limitation of the FOV as all the launch titles feature side by side full motion multiplayer when more than 1 person is playing.

Beyond that it may or may not be capable of say 3-4 people sitting with limited range of movement.

Regards,
SB