The lowest-level application layer is a signal processing system which
converts the raw sensor data into a collection of probabilistic
estimates. Conceptually, the classifier system allows the user to
train event recognition functions, or classifiers, to recognize
patterns in the sensory data and tag them as specific events. The
mechanism by which this time-series recognition occurs is a
multi-level HMM grammar which is capable of recognizing patterns at a
range of time-scales from seconds to days. A brief description of the
classifier system is provided below; for more information see the
Vismod technical report TR-519 available online at
http://vismod.www.media.mit.edu/tech-reports/TR-519/.