Database Reference
In-Depth Information
shower. The Gator Tech Smart House is adding healthcare technologies
to assist diabetes management.
A set of smart home environments called CASAS has been setup
in Washington State University. The CASAS home has five different
testbed environments. The first, referred to as Kyoto [100], is a two-
bedroom apartment that is equipped with motion sensors (positioned
on the ceiling 1 m apart throughout the space), sensors to provide am-
bient temperature readings, and custom-built analog sensors to provide
readings for hot water, cold water, and stove burner use. Voice over IP
captures phone usage, contact switch Q4 sensors monitor the open-closed
status of doors and cabinets, and pressure sensors monitor usage of key
items such as the medicine container, cooking tools, and telephone. The
second testbed, referred to as Cairo is a two-bedroom, two-story home.
There are three additional environments configured as single-resident
apartments (Bosch1, Bosch2, and Bosch3) that are part of a single as-
sisted care facility. All of these environments contain motion sensors
throughout the space as well as door contact sensors in key areas. Sen-
sor data for each of the environments are captured using a sensor net-
work and stored in a database. The data is analyzed for automatic ADL
recognition, monitoring of diabetic patient diet, and exercise adherence.
These environments also allow the presence of pets along with humans
to simulate realistic settings. Researchers employ Hidden Markov Mod-
els (HMMs) to recognize possibly interleaved activities from a stream
of sensor events, with the hidden states representing activities. There
is also strong emphasis on questions pertaining to the selection, place-
ment, and focus of sensors in a smart environment. In several studies
conducted by researchers [100], they have employed mutual information
(MI) based measures to rank sensors, and quantify the mutual depen-
dence between the sensor reading and the activity of interest. They then
use a filter-based sensor selection strategy to systematically evaluate the
effect of removing sensors with low MI values on activity recognition per-
formance. They also use hierarchical clustering to identify sensors with
overlaps in the field of view in order to remove unnecessary sensors, and
determine appropriate placements for the deployed sensors using a de-
cision tree learner. They have shown that reductions on average of 20
percent of the sensors are possible for different types of activities and
different configurations of the smart home.
Other examples of laboratory smart environments include a two-story
single-family house called Aware Home developed by the Georgia Insti-
tute of Technology. This is a living laboratory house designed primarily
to assist adults with cognitive impairment [98]. For instance, the home
includes a capture system on the kitchen countertop with a wall display
