none
Is there a way to discard the legs? RRS feed

  • Question

  • Hello all,

    I'm working on a project to enable patients in a wheelchair to play a game using their arms. The problem is that we want to let these patients sit behind a table. This way the legs are not visisble for the camera because the table covers them and as a result the arms are sometimes recognised as the legs and the hands are placed somewhere halfway the lower arm, making the recording very unreliable.

    Is there a way to 'tell' the system to stop attempting to find the legs? ideally only the torso and arms should be tracked

    If anyone has an idea how this problem could be solved i would be really grateful.

    Saturday, February 18, 2017 3:58 PM

Answers

  • As far as I know Kinect v2 tries to map the skeleton to you in seated mode as best as it can. If it can't do it properly then its probably due to lots of occlusion. If you occlude a human body too much the algorithm would probably have no way to map the whole skeleton. And there is no way to modify the skeleton mapping part.

    Do you need full upper body joint data or just the hand tips?

    If you just need the hand tips, you could try other ways like retroreflective markers(like Mocap) and try to track them yourself in the IR frame. You just need your patients to wear those markers and they should be able to play.

    • Marked as answer by EdBol Tuesday, February 28, 2017 11:16 AM
    Saturday, February 18, 2017 8:35 PM

All replies

  • As far as I know Kinect v2 tries to map the skeleton to you in seated mode as best as it can. If it can't do it properly then its probably due to lots of occlusion. If you occlude a human body too much the algorithm would probably have no way to map the whole skeleton. And there is no way to modify the skeleton mapping part.

    Do you need full upper body joint data or just the hand tips?

    If you just need the hand tips, you could try other ways like retroreflective markers(like Mocap) and try to track them yourself in the IR frame. You just need your patients to wear those markers and they should be able to play.

    • Marked as answer by EdBol Tuesday, February 28, 2017 11:16 AM
    Saturday, February 18, 2017 8:35 PM
  • Thanks for your reply.

    I actually need the whole upper body to be tracked, so i'm afraid markers on the hands won't cut it. Also the reason we initially chose a kinect as measurement device is because the patients need to be able to play the game in their home setting, so placing markers is not really an option.

    Sunday, February 19, 2017 9:46 AM
  • Does it have to be played on a table then?

    If not then you could probably try rotating both players towards the camera and they should be tracked in seating mode. Have a screen in front of them and have them wave in the air for interaction. You can also try and make the interactions in a way that allows them to keep their arms/hands down when they don't want to interact to avoid joint exhaustion.

    Monday, February 20, 2017 9:11 AM
  • The table will be mainly used to store some objects that are needed for the game. Part of the game is picking up objects and placing them on the right spot.

    I found out that the IR of the kinect can pass through glass, so my idea now is to use a glass table. I'm hoping the kinect will be able to recognise the legs through the table in this manner.

    While reading back this thread I realised I might not have been clear enough about the gaming part. It's not a multiplayer game, it's single player solely for rehabilitation purposes


    • Edited by EdBol Wednesday, February 22, 2017 1:49 PM
    Wednesday, February 22, 2017 1:44 PM
  • I don't think a glass table will be a good idea. Glass tables tend to be thick and there's usually refraction going on there and especially if you have the sensor face is at an angle or not adjoining the glass surface, the IR will probably never go back to the sensor. Therefore no tracking is possible. I might have been possible if you'd have the sensor at the table height but in front of the table so it could both over and under the table without obstructions and while the table still occludes part of the body it might not be a part important enough for the tracking algorithm to fail. But then you'd have trouble tracking the objects I guess.

    You could try it anyway but usually it matters. I've seen it many times on storefronts where the kinect needs to face towards the outside. Depends on the refraction index of the material.

    Wednesday, February 22, 2017 10:49 PM
  • I've tried it with a table of PMMA, and it looks pretty okay. The legs are not very accurately tracked but that is not a problem for us because we won't use any leg data. The legs are sufficiently tracked to stop the system from mixing up arms and legs so the arm and trunk data is much more accurate now.

    Thanks for the help, I will continue with the PMMA table for now. If it turns out the PMMA cannot be used, I'll report it here in case somebode else with similar issues stumbles on this thread.

    Tuesday, February 28, 2017 11:16 AM
  • Nice. Please do so. And good luck!
    Tuesday, February 28, 2017 1:06 PM