Kevin, we haven't tried a Kinect, so I don't have an answer for you, but I am interested in hearing more about your issue. Can you elaborate a bit on the problem? How do you think this variability in segment length comes about?

Thanks Sean . one of the obstacles in using the kinect for quantitative kinematics is that the body segments are dynamically determined and variable - unlike a optical motion capture system where markers are placed at segment endpoints. For gaming and identification of general body movements, I suppose its not important to have fixed segment lengths - I do not know how the kinect system does it. . I was wondering if anyone had a solution to use kinect for biomechanical analysis.

Okay, now I understand, thanks. We have contemplated using a Kinect for biomechanical measurement, but decided that its spatial accuracy is not sufficient for certain movements during gait. This article provides some info:

http://www.mdpi.com/1424-8220/12/2/1437/htm

It seems that the sensor would have to be extremely close to the subject to get similar accuracies to marker-based motion capture. We never even knew about the issue regarding segment lengths. Thanks for posting this.

I'm also very interested: i'm cannot find any free software to extract the relevant coordinates. i found some software here: http://tech.integrate.biz/kinect_mocap_instructions.htm and a relevant abstract herehttp://www.degruyter.com/view/j/bmte.2012.57.issue-s1-F/bmt-2012-4452/bmt-2012-4452.xml. could anyone provide me with the relevant software? We want to use the kinect motion analysis system for educational purposes in a bachelor physiotherapy to discuss the topics reliability and validity

I'm using the Kinect for motion tracking of the upper extremities. The Kinect SDK will give you the skeleton joint coordinates and then you can just do some simple vector math to calculate the joint angles. Clearly it isn't going meet $100,000 motion tracking system standards but I have been pleasantly surprised by the accuracy.

Thanks Elizabeth, That is our objective , yes we do get kinematics easily but we are working on algorithms to improve the precision to match other optical systems , and one of the issues is the dynamic segment length issue . I was curious to see if/how others had solutions.

Kevin

Hi Kevin and Elisabeth,

I am very curious About your software systems. Is it available for download somewhere? THE first link is posted did give me some software, but THE data it presents are not coordinates, so i can not do THE vector calculations.

Kevin, the link i posted was about applying the levenberg marquardt algorithm to freeze segment lengths. Since i am not into ICT I don't know if this is easy to integrate.

Again, I'm very interested, so if anybody could send their (preliminar) software I would be gratefull.

Jaap

Jaap, We are not using any special "software Systems" . We are just using the SDK as Elizabeth mentioned above to get coordinates right now .

Hi everybody,

we used the Kinect for some experiments. I think by now the SDK from MS and the OpenSource solutions are somewhat limited: For example, both assume, that the actor is facing to the kinect. If you turn by 180degrees this is usually not recognized. Thus, I'm not sure if you can get a motion capture of high quality from these systems. I saw recent publications on this topic:

http://faculty.cs.tamu.edu/jchai/projects/sigrapha-asia-12/body-capture/a188-wei.pdf (SIGGRAPH Asia 2012)

and

http://www.mpi-inf.mpg.de/~abaak/download/2011_BaakMuBhSeTh_DataDrivenDepthTracking_ICCV.pdf (ICCV 2011)

But as far as I know the software is not available for download so far.

Best regards!

Dear all

I am very interested in cheap devices for motion recording such as Kinect. Although I still never used the device I was collecting information about and, beside the SDK, I found software developed in “LABVIEW” which could be also considered to extract and analyze data from Kinect. A lot of documentation may be found simply typing the keywords in any search engine and specific packages may be downloaded at

http://sine.ni.com/nips/cds/view/p/lang/it/nid/210938

http://labviewhacker.com/doku.php?id=%20[[projects:lv_kinect_interface:lv_kinect_interface|]]

Fabio

Hello every body

Sophisticated, and of course expensive, motion analysis systems are useful just in the labs. Sure the research in the labs can lead into better understanding of the movement disorders and possible solutions, but it has never been a clinically easy task. Therefore, any system reasonably priced and handy enough to be used in a clinical level with a descent level of accuracy will be a real help in this regard. I have been and still am working with some motion analysis systems and considering my physiotherapy background, I am really interested in working with anybody who would like to move this agenda to the level of the clinics and real patients.

Hossein,

We have spent the last few years delving into this problem of clinical motion capture as compared to research motion capture. You are absolutely correct that there is a need for more accessible, yet accurate, methods of motion capture. As a potential solution, we currently offer a turnkey motion capture system for gait analysis that utilizes off-the-shelf cameras with custom analysis software.

http://3dgaitanalysis.com/

Sean

Hi Sean, Thank you very much for the information. It looks pretty good. I will dig further into it and may get back to you for help/advice.

Everyone should check out this video of the new Kinect sensor, which was announced yesterday. In addition to other features shown in the video, the sensor has much higher resolution and the default skeleton tracking now provides wrist supination/ pronation (YAY!) and gross grasp.

http://www.youtube.com/watch?v=Hi5kMNfgDS4

It would be very interesting to see how this compares with more high tech motion capture. I would be interested in any thoughts. Could be of great clinical use if it can pick up clinically important differences.

Hi everyone, I am currently developing an application using MATLAB + image aquistion toolbox. You get both the depth image and the RGB images.

My undergrad dissertation was on using the Kinect for human motion analysis. Unfortunately due to both dynamic segments and hence inconsistent coordinate systems alongside inaccurate pose estimation the default skeleton tracking was not very useful. Using OpenNI and and some open-source Matlab .mex files you can easily grab the raw 3D point cloud and manually take 3D measurements, or attach black markers and use simple marker-tracking code to obtain 3D marker positions.

In terms of accuracy, the depth data is quantised to 11-bit in a way that favours precision at short distances. So the quantisation step increases (and hence precision decreases) with increasing distance. The raw data is actually of much greater resolution however has to be downsampled due to the limited bandwidth of USB2.0.

Other than waiting for the Kinect 2, there is another device called Leap Motion which is about the be released and claims far greater accuracy than the original Kinect.

I just uploaded my dissertation in case it is useful to anybody.

https://www.researchgate.net/publication/244988901_Can_gaming_technology_be_used_for_human_motion_analysis?ev=prf_pub

Thesis Can gaming technology be used for human motion analysis?

My old colleagues at the Centre for Sports Engineering Research at Sheffield Hallam University have been doing quite a bit of work in this area. Check out: http://www.depthbiomechanics.co.uk/. A paper was also presented at ISBS 2012 and a review paper in Sports Technology is due shortly.

Our group has been actively pursuing research in this area, specifically, comparing quantitative analysis using high-end mocap systems and new-age COTS systems like kinect. Check the following list of publications (entire list available in the lab website: http://mechatronics.eng.buffalo.edu/publications) -- since, the PDFs are not yet updated on the webpage, let me know if you are interested in details of any of these.

Thanks

Madu

+++++++++++++

[C78] Jun, S.K, Zhou, X., Ramsey, D., and Krovi, V., “Automation for Individualization of Kinect-based Quantitative Progressive Exercise Regimen,” 2013 IEEE Conference on Automation Science and Engineering Wisconsin, MI, August 17-21 2013.

[C77] Jun, S.K, Zhou, X., Ramsey, D., and Krovi, V., “Kinetostatic Design-Refinement of Articulated Knee Braces,” Proceedings of the ASME 2013 International Design Engineering Technical Conferences and Computers in Engineering Conference, ASME IDETC/CIE 2013, August 4-7, 2013, Portland, OR.

[C71] Jun, S.K, Zhou, X., Ramsey, D., and Krovi, V., “A Comparative Study of Human Motion Capture and Analysis Tools,” 2013 International Conference on Rehabilitation Robotics, Under Review, June 24-26 2013, Seattle, Washington USA.

+++++++++++

Here's the Sports Technology paper I was referring to: http://www.tandfonline.com/doi/full/10.1080/19346182.2013.819008#.Uhc2CH9y3-U

Nice discussion, data and info about Kinect! Thank you.

MATLAB R2013a y R2013b comes with Kinect toolbox. Has anyone try it to collect data?

Hi all,

Does anyone know any ways how to capture, and if yes with what software and with what accuracy head rotation and tilt with Kinect? thanks

Presentation software (www.neurobs.com) has added a Kinect 1.0 interface that integrates body position monitoring with stimulus delivery. On a relatively fast multi-core PC, Presentation can control an experiment while storing the video data, depth data, and joint position data in C3D format. A 30s video can be seen here: www.neurobs.com/menu_presentation/menu_features/kinect

Measurements with the Kinect 1.0 are imprecise. The Kinect 2.0 sensor is being tested and should improve performance. It will be integrated with Presentation in early 2015. Head rotation tracking will also be added.

We have recently used the kinect to study Karate movements. You can find the paper at: http://www.mltj.org/index.php?PAGE=articolo_dett&ID_ISSUE=765&id_article=6564

Gonçalves, N., Rodrigues, J. L., Costa, S., & Soares, F. (2012, September). Automatic detection of stereotyped hand flapping movements: two different approaches. In RO-MAN, 2012 IEEE (pp. 392-397). IEEE.

They use Kinect. 

Hello guys

I am trying to capture the movement of an array of markers on top of my soft robot using Kinect v2. Can anyone help me with it ? Is it possible to do it with Kinect v2?

Dipankar Bhattacharya, you might want to consider using a mocap system like Vicon Nexus rather than the Kinect for tracking markers. You might be able to do it with the Kinect, however, this requires substantial programming (time consuming!) and would still be less accurate than the Vicon system.

Thanks Lukas. I am already using Vicon system for tracking the markers. I wanted to design a similar kind of mo cap system by using kinect and wanted to compare its results with the Vicon Nexus. Thanks again for replying.

Hello..which software are you using with your KINECT 2 sensor for motion analysis???@