none
Output skeletal joint data RRS feed

  • Question

  • Hi I am a noob.  I need to output the skeletal joint coordinate data either in real time or as a file post data acquisition.  We are trying to track the motion of newborn infants. 

     

    Also does anyone know how small of a skeleton the kinect hardware is able to track out of the box? 

     

    Thank you.

    Monday, July 18, 2011 8:51 PM

Answers

  • AAARRRSINIC,

    You should take a look at SkeletalViewer sample that gets installed with SDK (C:\Users\Public\Documents\Microsoft Research KinectSDK Samples\NUI\SkeletalViewer), available in both C# and C++. Also look at documentation and walkthroughs available in http://research.microsoft.com/en-us/um/redmond/projects/kinectsdk/guides.aspx

    As far as outputting data to file for later analysis you'll need to write that code yourself, since the Kinect SDK Beta doesn't provide functionality to save this data on your behalf.

    Note that the skeleton tracking algorithms haven't been extensively tested for infant posture, but if all limbs of infant are clearly visible by camera you should be able to get reasonable results.

    Eddy


    I'm here to help
    Monday, July 18, 2011 9:43 PM

All replies

  • AAARRRSINIC,

    You should take a look at SkeletalViewer sample that gets installed with SDK (C:\Users\Public\Documents\Microsoft Research KinectSDK Samples\NUI\SkeletalViewer), available in both C# and C++. Also look at documentation and walkthroughs available in http://research.microsoft.com/en-us/um/redmond/projects/kinectsdk/guides.aspx

    As far as outputting data to file for later analysis you'll need to write that code yourself, since the Kinect SDK Beta doesn't provide functionality to save this data on your behalf.

    Note that the skeleton tracking algorithms haven't been extensively tested for infant posture, but if all limbs of infant are clearly visible by camera you should be able to get reasonable results.

    Eddy


    I'm here to help
    Monday, July 18, 2011 9:43 PM
  • Thank you Eddy,

     

    We will certainly position the camera in a way that the infant is clearly visible by the camera.

     

    If anyone on this forum has already written code in C# for this and is willing to share it with me, it would be GREATLY aprpeciated.

     

    Thank you.

    • Proposed as answer by B. K. Brown Wednesday, July 20, 2011 4:05 AM
    • Unproposed as answer by B. K. Brown Wednesday, July 20, 2011 4:06 AM
    Tuesday, July 19, 2011 6:50 PM
  • I recently modified the SkeletonViewer example to map left-hand XY coordinates to a USB-based data acquisition module to control an inexpensive remote control boat in real-time. I had to add some glue electronics to interface the DAC module to the boat's modified RC handheld controller. There are a few videos (unfortunately poor quality) posted here:

    photobucket.com/bryansbrain

    Is this in line with what you have in mind? If so I could send some pointers.

    Wednesday, July 20, 2011 4:14 AM
  • I am curious as to how you would position the kinect and infant such that the kinect is able to detect the infant.
    Monday, July 25, 2011 1:46 PM
  •  

    How adorable is this?? Being an expectant mom and currently working a lot with the kinect I am really curious to what you do with this concept!  Please let us know how your project progresses!

     

    If it helps - I have a toddler mannequin here that is 31 inches tall and skeletonviewer picked up the skeleton OK. 

     


    Susan
    Monday, August 1, 2011 12:20 PM
  • Hi Susan,

    we also tried to do skeleton tracking with a mannequin, but the kinect is not detecting depth data of mannequing belonging to a player.
    Which material has your mannequin?

    Regards

    RB

    Friday, June 8, 2012 11:46 AM
  • Hi RB - We've worked with several different kind of mannequin, these include some kind of plastic (like a department store mannequin), resin (body molded from an actual person), cloth (clothing form mannequins) and a mannequin that we call "nerf boy" who is made of something similar to 'Nerf" material. 

    We've been able to get 3D data from all of them - skeleton tracking is a different story.  I believe that mannequins cause problems with the skeleton tracking because they stay absolutely still. When working with a mannequin I found I had to jostle it a bit to get the skeleton detection to 'snap in'. 

    I think you're having that same problem.  It'll see the 3D data of the mannequin but not recognize it as a player - that's because it hasn't detected a skeleton for it.   I'd guess either the mannequin isn't sufficiently recognizable as 'human' or that you need to move it around a bit in the view. 

    Hope this helps!

    Susan


    Susan

    Friday, June 8, 2012 12:13 PM
  • Hello Susan,

    thank you for your answer. We are getting depth data from the mannequine, but no skeleton data.

    The mannequine is pretty realistic wrt human shape.

    We need the skeleton data. We know the problem of non-moving players and we have to introduce some small initial movement.
    The first try to circumvent that problem was to tilt the camera up and down to

    introduce an artificial movement, but the kinect is turning off the IR emitter while tilting, so that didn't help.
    Moving the mannequine manually also didn't really help. As soon as i entered the view of the kinect

    the kinect recognizes me and when I touch the mannequine and move the dummy, the mannequine data is part

    of my data, but recognized. As soon as I take of hands from the mannequin the Kinect is not detecting

    the mannequine as a player any more
    I am using SDK 1.5 now. Have you tried it or are you still on 1.0?

    Regards

    RB

    Friday, June 8, 2012 12:30 PM
  • We've moved to mainly real people testing now so it has been some time since I've worked with the mannequins.  We're using the latest commercial released SDK now (I forget what that is offhand).  

    I understand what you're saying about the kinect seeing you when you go into the scene to move the mannequin, I had the same trouble.  As a result I wound up getting a lot exercise in my last couple months of pregnancy getting up and down from my chair to test, because the mannequin was not reliable!!

    I would suggest either wiggling the kinect manually or wiggling the mannequin with some rigging without going into the scene.  But I don't have a surefire answer, sorry!

    Susan


    Susan

    Friday, June 8, 2012 12:47 PM
  • Hello Susan,

    thank you again for your anser.

    Just for your information. It also didn't help to wiggle the cam around.

    The Kinect seems to be blind for our mannequine, but based on your anser, it seems that there are

    mannequines which work fine.

    So we have to find one that works I guess.

    Thank you

    RB

    Friday, June 8, 2012 1:03 PM