This week was devoted to analysis of the data that I collected last week. First, we took simple means and standard deviations of the subjects' completion times. The hard part is cropping up in the analysis of the hands' motion data. We want to look at a user's acceleration profile while grabbing both real and virtual objects. In order to do this, we need to look at the hand's position data and extrapolate when a subject grabbed an object. I did this yesterday by looking at the x, y, and z coordinates separately; I categorized a grasp by a somewhat steady position in the z direction, a downward movement in the y direction and a move forward and then backward in the x direction. This was all done by hand (no pun intended!) and 6 grasps were separated from the data for each participant, 3 in the virtual environment and 3 in the real world.
The next step will be to take this data and plot the acceleration profiles of each participant. The acceleration profiles of a human in the real world when grabbing an object is well documented and shown to resemble a bell curve, speeding up in beginning, and then slowing down as the hand approaches the object. It will be interesting to see how closely the acceleration profiles from the virtual world match those in the real world. If they are similar, we can conclude that perhaps the interface is a good first step towards providing effective virtual object manipulation. We always want virtual world performance to mirror real world behavior as closely as possible, and making sure that our natural movements are not being altered by a virtual environment is important to achieving this goal. If the acceleration profiles differ greatly then we can begin to look at why this is and try to fix this problem. For example, implementing Python's physics engine would ensure that users couldn't simply stick their hand through an object and must be more accurate when approaching the location of the object.
Before any conjecturing can be done, however, I have to finish with the analysis and this means filtering the noise out of the data and looking at the results.
