View Full Version : Summary of responses to How to reconstruct the position of themissed markers during human motion capture

02-21-2007, 03:19 AM
Dear all,

Many thanks to all of you who responded to my query regarding the reconstruction of the missing marker spatial trajectory during human movement capture.

I have posted my original query followed by a summary of responses. Hope other people find the information below as useful as I did.

Thanks again,

Feng Yang

************************************************** *******************
Original query
The motion-capture systems (such as Vicon, Qualysis, and Motion Analysis, etc.) have being widely used in human movement analysis. It could be frequently encountered that some markers are missed for several consecutive frames during motion capture. Almost all commercial systems mentioned above have the function to fill the gap of the missed marker's trajectory by using cubic spline or linear interpolation. However, the accuracy is not high enough using this function. Sometime, the filled trajectory is totally wrong.

Does any one have same experience to solve this problem? Is there any reference which describes the reconstruction of the spatial trajectory of the missed marker?

************************************************** **********************

Summary of responses

1. Using the commercial software
Visual 3D
Virtual join (Motion Analysis)
Copy Pattern (Vicon)
The ISB website (www.isbweb.org) has spline fits for just this purpose in their SOFTWARE RESOURCES section, specifically Woltring's Generalized Cross-Variance (GCV) Natural B-spline Filter by Tony Reina. The filter is available for both C++ and Matlab.(James D. Johnston, MSc).

2. Self-developed mathematical methods
a) If you have a rigid body or if that is a valid assumption, then the distance between markers on the rigid body will remain the same. Therefore you can reconstruct the position that the marker must be in using transformation matrices. I think you may need 4 markers per segment for that or you will have to have some other orientation constraint. (Marilyn Powers, Kjartan Halvorsen)

b) Treat the x, y and z streams of the missing marker independently. Generate the plot of single stream against time to create a cyclic graph, then isolate the problem areas and replace using a straight line fit that matches the mathematical pattern of single stream track. (Steve Abbott)

3. Adjust the position/orientation of the cameras, organize the capture environment to improve the capture quality.

Also, several references were provided as follows.

Fradet, L. et al. (2004). Do handball throws always exhibit a proximal-to-distal segmental sequence? Journal of Sports Sciences, 22, 439-447

Signal processing in movement analysis (a state-space approach) Human Movement Science, Volume 15, Issue 3, June 1996, Pages 389-410

Halvorsen K, Soderstrom T, Stokes V, Lanshammar H.Using an extended kalman filter for rigid body pose estimation.J Biomech Eng. 2005 Jun; 127(3):475-83.

Finding fabulous fares is fun.
Let Yahoo! FareChase search your favorite travel sites to find flight and hotel bargains.