dingvale
commented
9 years ago Notes from research on decision tree: Need to find factors from data to use as decision-making nodes in tree, i.e. not just mean & variance, but perhaps some way to quantify "frequency" of movement. If speed is an issue, we need to find a way to balance number of decision nodes and classification accuracy, which can be achieved through testing.
Input: features extracted from wearable sensor data, first by emulating stereotypies characteristic of ASD, then (possibly field data, but not counting on it). Output: quantification & classification of the movements and matching to possible characteristics. Nicola brought up the point that it can be hard to tell whether or not a motion is a stereotypy, or actually a typical method to reduce stress/anxiety.