Improving user comparison

[michael5486]

Joint tracking has been improved, working on having the program continuously update it's knowledge about the user...this is the beginnings of the "machine learning" aspect

[michael5486]

Can get the median of the torsoHeight...using the median offsets the impact of outliers, and after a couple seconds there is a pretty reliable distance measurement for torso

[michael5486]

changing calcDistance function to use imageX and imageY coordinates, and using the median worldZ value. Needed to get accurate distance measurements between joints because the raw world coordinates are unreliable, but the image coordinates are reliable

[michael5486]

calcDistance fixed

[michael5486]

Made a lot of progress with user last known location (LKL). Program keeps a buffer of recent locations for the user and uses that in another comparison if the target user cannot be found.

[michael5486]

LKL needs some refining to assist in user permanence...trying a new algorithm to divide the image into regions and filter out users in far away regions from where the TU was last seen

[michael5486]

Still working on improving user comparison, refining thresholds and such with further experimentation.