Announcement

Collapse
No announcement yet.

Motion capturing using Kinect

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Motion capturing using Kinect

    Hi,

    I am interested in knowing whether anyone has any experience in using Kinect for motion capturing?
    Has anyone developed any software package to use the data from Kinect to do kinematic or even kinetic analysis? Can we use Kinect instead of other motion capturing systems like Vicon? I have done research with Vicon and I know its capabilities, but if Kinect can provide some meaningful results, it will very interesting in both fields of biomechanics and robotics.

    Thanks in advance

  • #2
    Re: Motion capturing using Kinect

    Babak,

    we had a close look at the measurement features of the Kinect product.
    At first the capabilities look promising, but when you look closer the deficiencies become clear:
    • ~ 30 frames per second are not enough for most movements.
    • for a full body view you need a distance of about 3 m, resulting in 2.8 cm resolution in depth.


    You may find a solution with multiple Kinects (they are cheap), but this will take quite some efforts to compute the combined results.

    For the moment the Kinect is a toy, but version updates and competitors will show up within reasonable time, so one day it might be possible...

    Regards,

    Thomas.

    Comment


    • #3
      Re: Motion capturing using Kinect

      I would recommend a search on this very form for the term "kinect". You'll find a couple of topics about using the kinect for biomechanics research. Two postings that stand out in my memory are these:

      Comment


      • #4
        Re: Motion capturing using Kinect

        Thank you for the links that you have suggested. I already have looked at them. It seems that they didn't peruse their research on Kinect.

        Comment


        • #5
          Re: Motion capturing using Kinect

          Hi There,
          We have continued the accuracy study with a number of different motions and participants and also using the IPI soft motion capture software. Unfortunately the time we have to spend pushing this forward is limited, although we hope to publish something soon! I am presented a limited section of the data at this years ISBS conference in Melbourne, the papers may be available to read after the conference.

          All the best,

          Simon Choppin

          Comment


          • #6
            Re: Motion capturing using Kinect

            Originally posted by tseeholzer21 View Post
            Babak,

            we had a close look at the measurement features of the Kinect product.
            At first the capabilities look promising, but when you look closer the deficiencies become clear:
            • ~ 30 frames per second are not enough for most movements.
            • for a full body view you need a distance of about 3 m, resulting in 2.8 cm resolution in depth.


            You may find a solution with multiple Kinects (they are cheap), but this will take quite some efforts to compute the combined results.

            For the moment the Kinect is a toy, but version updates and competitors will show up within reasonable time, so one day it might be possible...

            Regards,

            Thomas.
            I would definitely agree with this summary.

            We have performed a within-session and between-session reliability study with 5 subject performing a slow squatting task and the results are not good enough for us to contemplate using a kinect for even basic clinical motion capture.

            We are also in the process of preparing a manuscript for publication - hopefully it will add to the debate.

            At this stage though its best to wait for a better spec'ed device to arrive on the market.

            Corey

            Comment


            • #7
              Re: Motion capturing using Kinect

              This is a most interesting discussion and be an example of another development that may increase the opportunities for motion capture to be perfromed in many more contexts. I have a colleague in gaming at my university who is also pursuing the usefulness of the Kinnect system, so I would be keen to hear what others have found so far too. Simon, I will be at teh ISBS this year so it would be good to catch up with you about this and all tne best with the publication Corey.

              Comment


              • #8
                Re: Motion capturing using Kinect

                Originally posted by tseeholzer21 View Post
                Babak,
                for a full body view you need a distance of about 3 m, resulting in 2.8 cm resolution in depth.
                Regards,
                Thomas.
                Hi Thomas, Do you know the relationship between resolution and distance?

                Regards,
                Scott

                Comment


                • #9
                  Re: Motion capturing using Kinect

                  Originally posted by smcmillan53 View Post
                  Hi Thomas, Do you know the relationship between resolution and distance?
                  Obviously the resolution in depth gets worse with increasing distance.
                  These are the numbers we found :

                  1.0 m :: 0.3 cm
                  2.0 m :: 1.2 cm
                  3.0 m :: 2.8 cm
                  3.5 m :: 3.9 cm

                  The distance in [cm] is the difference between two adjacent measurement results (1 bit difference).

                  Regards,

                  Thomas.

                  Comment


                  • #10
                    Re: Motion capturing using Kinect

                    Even though the Kinect is not nearly as accurate as a Vicon Motion Capture System (VMCS), there are still some pretty good qualities that can be utilized, which we think might make it worthwhile considering the significant price comparison. Recently, we've done some work with tracking upper-limb movements and determining specific kinematics (i.e. Range of Motion, Peak Angular Velocity, etc) using the Kinect and comparing the results to the same data captured by a VMCS. Some of the preliminary results will be published in BIOROB 2012 (look for Authors: Douglas Brooks, Yu-ping Chen, and Ayanna M. Howard). There's still a few areas that we're working to resolve (i.e. real-time analysis, gait-kinematics, etc.), but for a low-cost system (that can easily be transported), it shows viability as a first-step before a more in-depth evaluation is performed.

                    Comment


                    • #11
                      Re: Motion capturing using Kinect

                      I just came across this paper which might be of interest
                      Dutta, T. (2012). Evaluation of the Kinect™ sensor for 3-D kinematic measurement in the workplace. Applied Ergonomics, 43(4), 645-649. doi: 10.1016/j.apergo.2011.09.011
                      Justin

                      Comment


                      • #12
                        Re: Motion capturing using Kinect

                        At Simi we did a lot of research and development for markerless tracking methods within the last 3 years, also including a government research project. We also tested the Kinect cameras in a master thesis and found the same results as others reported before. Biomechanical the results are not sufficient.

                        We are just testing our first release of a markerless tracking system that will be integrated with Simi Motion. It is based on silhouette tracking. Preliminary results look good in terms of practicability and accuracy. We expect the first version to be available within the first half of this year. If you are interested to get some more information about the new system please write to me.

                        Comment

                        Working...
                        X