Can somebody explain the process behind obtaining ground truth values from a motion capture device, which can be used for comparison with our proposed system like for eg: Kinect? tell me whether we have to do our work in the presence of both the 3D motion capture (ground truth) and our proposed system(Kinect)
Hello! I will look for an article that establishes some standards of measurement through 3D video capture , specifically oriented gait analysis , when I have access to the scientific article I will send you the link
regards !
Optical motion capturing is often considered to be the 'gold standard'. If you develop another technique (can be Kinect, accelerometers, video based) to capture motion, you can express the accuracy in terms of how close you are to motion capturing. To this end, you have to capture a few sequences with both systems simultaneously and do your evaluation on this sequences.
Note, on this basis you can't come up with a technique that is better than the gold standard, since you don't have a reference to proof that.
Two examples of alternative techniques that are compared against motion capture can be found in these papers (and more in the references within):
Accelerometer based:
http://cg.cs.uni-bonn.de/en/publications/paper-details/tautges-2011-MotionReconstruction/
Image based:
http://cg.cs.uni-bonn.de/en/publications/paper-details/yasin-2016/
Best!
Hi everyone
As per the standards of 3d no equipment is gold standard, there are false positive results and give under rated or over rated value
Hi All,
Like what Senthil is saying, there is no real gold standard but the closest to that would be motion capture systems such as Vicon or optotrak. To compare your method you will need to capture the same motion using one gold standard and your technique. In addition you can run these motions multiple times to test repeatability. Then you can get the RMS error or standard deviation as measures of how tour technique measures with these gold standards. Your technique could match the accuracy of gold standards but offer measurement outside structured environments.
The most accurate motion capture systems are the marker and camera systems like Vicon or OptiTrack. With these systems, you can attach several markers to the human body and track the movements. The errors obtained by the mentioned systems are low relative to the other motion capturing methods.
There is also another way of comparison; you can attach several precise inertial sensors to body segments, and by the orientations obtained by the inertial sensors you can compare with the Kinect's results.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling