none
Modifying the datastream to track skeleton RRS feed

  • Question

  • I am curious if it is possible to modify the datastream such as ColorStream or DepthStream such that the datastream reads color or depth images stored beforehand. It will be helpful for skeleton tracking on user specified images instead of streaming directly from Kinect. I think it is technically possible because Kinect Studio could also play back the recorded stream but the xed format is not defined in public now or I couldn't find it. Any suggestion would be helpful. 
    Wednesday, June 26, 2013 6:50 AM

Answers

  • I can stored colorStream and depthStream frames as those are only array of bytes. You can also record skeletonStream. (it's exactly what Kinect Studio is doing)

    Sadly developers can't inject recorded Color/Depth streams into Kinect API in order to get generate a delayed skeleton tracking.

    We may be luckier with the new Kinect ;)


    Vincent Guigui Innovative Technologies Expert at OCTO Technology MVP Kinect

    Wednesday, June 26, 2013 10:50 AM

All replies

  • I can stored colorStream and depthStream frames as those are only array of bytes. You can also record skeletonStream. (it's exactly what Kinect Studio is doing)

    Sadly developers can't inject recorded Color/Depth streams into Kinect API in order to get generate a delayed skeleton tracking.

    We may be luckier with the new Kinect ;)


    Vincent Guigui Innovative Technologies Expert at OCTO Technology MVP Kinect

    Wednesday, June 26, 2013 10:50 AM
  • Thanks for the quick reply. I just tried to use Kinect Studio to record Color and Depth streams. Then, I played back the two stream in the example "Skeleton Basics" in which the skeleton showed up. It looks like the skeleton can be computed based on the recorded Color and Depth Stream. In other words, if I can generate Color and Depth Stream from color and depth images from other sensosrs such as Asus Xtion Pro, would it be possible to generate the skeleton based on the images captured from other sensors? 
    Wednesday, June 26, 2013 6:40 PM
  • Yup I wonder too, technically it shouldn't be difficult but just Microsoft didn't release API for that... I want skeletal tracking to disable some player, if we can pass modified data to skeletal tracking, can just erase the corresponding depth data
    Friday, June 28, 2013 7:06 AM
  • We do not expose a way to do this from the Kinect for Windows SDK. The data generated by the runtime is passed to the application layer only. It has been requested as a feature.

    At this time, the only supported method would be to record and playback with KinectStudio.


    Carmine Si - MSFT

    Tuesday, July 9, 2013 5:52 PM