Figure 1.

Overview of the proposed algorithmic framework for data quality control of sensor data for behavioral clinimetric testing. The first stage (starting from the left) involves collecting sensor data outside the lab. The second stage consists of removing confounding factors from the data such as the effect of the orientation of the device. The third stage involves unsupervised segmentation of the data into intervals in which the user performs similar activities. In the final stage, a simple, interpretable classifier is trained to predict which intervals are associated with adherence to, and which with violations of, the test protocols.