I've been looking into this for a while now, but I don't think I know enough of the terminology to phrase this well enough for Google (my apologies).

So essentially I'm looking at motion capture data (multidimensional arrays set up as [frame number, point number, x coordinate, y coordinate, z coordinate]). Unfortunately when trying to compare motions, I'm unsure of how to move forward. I'm thinking something like PCA, but because there aren't always the same number of frames, I'm having a bit of trouble (the arrays are jagged).

I'm doing this work in Python, but I'm genuinely curious from a statistical point of view--is there anyway to do multidimensional analysis like PCA with jagged data inputs? Is that statistically unsound for further analysis? Is there anyway to transform the jagged data to better fit another type of analysis?

  • $\begingroup$ What would be the goal of such an analysis? $\endgroup$ – kjetil b halvorsen Mar 31 '17 at 19:49
  • $\begingroup$ I'm looking into comparing differences between particular motion captures (so that they can be classified, for example, or compared for global/local similarity). I'm sorry that's​ so vague, but I want to be able to essentially create a system of scoring motion sequence "hits" in a similar methodology to BLAST for protein or gene sequences (which is also hard because BLAST data is one-dimensional). $\endgroup$ – user3684314 Mar 31 '17 at 20:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.