In the morning today, I finished organizing my code to make it easier to add functions to clean the data and remove errors. After that, I worked on editing my code so that it would interpolate values for the vectors that previously had invalid x, y, z values (for example, a vector with a -z coordinate is invalid since having a -z coordinate implies that the eye was looking inside the head). The first function that I was using to interpolate the values did not work because it created some vectors that had a z value of zero (which implies that the eye is not staring into the 3D space at all). Then I found another function that I could import from SciPy package that would help me produce the results I needed. After editing my code, I tried to run it but I ran into an import error (the PyCharm IDE could not find the SciPy package). This will hopefully be resolved tomorrow and I will be able to test if the functions work and if they clean the data successfully.
In between work on our projects, Gabe and Kamran demonstrated the walking study that they had been working on this past year. In that study, they recorded the feet and leg movements of participants (using motion capture) as they walked over both physical and augmented reality (AR) obstacles. The purpose of their study was to learn whether there is a significant difference in how people react to AR obstacles when walking versus physical obstacles in the real world. The results of their study showed that there was not a significant difference between how people react to AR obstacles and physical (real world) obstacles. These results support the idea that research testing people's reaction to certain scenarios can be conducted in AR environments and have accurate results.
This image below depicts two different types of interpolation (linear and cubic). Linear interpolation draws a line between the consecutive data points and cubic interpolation fits a cubic function to the data. When interpolating the data for the eye tracking, I will try and see whether linear, cubic, or higher-power polynomials fit the data best.