【Experiment】Generate slope: 02-01-2013

In this experiment, I generate the slope values to model the temporal variation of motion data based on the slope difference. The data is 20 taichi motions, which are collected by Mao, Ma, Li. Here are the steps:

1 Exteract the warping functions between paire wise motions. The purpose is to extract such temporal variation information that can be applied to other motions to generate temporal variation.

2 Extract the slope values based on the warping function for each segment. 

-------------------------------------------------------------

   mean for the data(slopealltogether): 0.9478

   std for the data(slopealltogether): 0.3450

 

min(slopealltogether)=0.2340
max(slopealltogether)=2.5909

-------------------------------------------------------------

3 Model the slope value distribution using Log-Normal distribution.

   log-normal:
   parameter =

-0.1228 0.3834

4 Model the relation between previous slope value and the slope difference (current-previous) y=ax+N, where x is previous slope values, and y is the slope difference, and N is Gaussian distribution, N(0,std)

--------------slopeXYdiffer-----------------------

mean(slopeXYdiffer)=0.0239
std(slopeXYdiffer)=0.3579

min((slopeXYdiffer))=-1.5732
max((slopeXYdiffer))=1.3257

 

standeviation=sqrt(sum(error2)/N)= 0.2825; % For N, the gaussian distribution

p=polyfit(testX,testY,1);
p = -0.6260 0.6140;

y=-0.6260*x+0.6140+N

--------------------------------------------------

 

4 Then generate slope based on this log-normal distribution, and for the next slope vaule