02-10-2012, 02:24 PM
Wireless Sensor Networking of Everyday Objects in a Smart Home Environment
Wireless Sensor Networking.pdf (Size: 163.04 KB / Downloads: 259)
Abstract
Within a smart home environment the information
processing is supposed to be thoroughly integrated into everyday
objects. This introduces the need to keep track of the everyday
objects and their state changes produced based on the user’s
interaction with them. Such information is useful in recognizing
the user’s activities, situations, etc. In this paper we present a
ZigBee communication protocol based wireless sensor
networking of 42 everyday objects (embedded with 81 simple
state change sensors of 8 sensor types) in a living laboratory
smart home environment. The system was evaluated in a
realistic setup with background noise. The sensing module has
shown promising results with an overall system precision of 91.2
% and a recall of 98.8 % in keeping track of the state changes to
everyday objects. The signal strength measure above the
acceptable limit of >10 dB to obtain reliable data communication
was found to be 97.5% checked at 8 different locations in a home
environment. Finally the transmission-reception range was
evaluated to be 33 m with a single wall obstruction and 19 m
with multiple wall obstruction in an indoor environment.
INTRODUCTION
Within a smart home environment the information
processing is supposed to be thoroughly integrated into
everyday objects that provide functionalities beyond their
primary purpose, there by enhancing their characteristics,
properties and abilities [1]. By correlating the sensor output of
such everyday objects, the wireless sensor network (WSN) as
a whole can potentially provide functionality that an
individual everyday object cannot. Such functionalities
include situation and activity awareness (using a middleware)
of an inhabitant within smart home environments. The
everyday objects that are present in a user’s environment have
shown to provide valuable cues for inferring the user’s
current situation and activity [2]. There are several wireless
micro sensor motes including Mica2Dot [3], iMotes [4],
BTNodes [5], Smart-Its [6], Smart-Its Particles [7], etc.
available to the research community. Even though many of
such motes have their own advantages, they do not meet the
general requirements of a smart home environment in sensing
the state changes to everyday objects including home
appliances, furniture, simple objects, etc. [8]. The ease of
installation.
SYSTEM OVERVIEW
The system to be described in this paper consists of a set
of everyday objects present in a smart home environment
connected to a wearable personal server [13] worn by the user
and running an activity-centered computing middleware [14].
A. Wireless Personal Area Network
The everyday objects are embedded with stick-on nodes
that sense the internal states and state changes (based on the
user’s interaction with it) to the objects and transmit this
information wirelessly using ZigBee communication protocol
to the user’s personal server. ZigBee was preferred over
Bluetooth for wireless personal area networking considering
its usage of low-power digital radios intended for low data
rate, long battery life and secure networking applications.
ZigBee supports up to 65,000 nodes on a network,
introducing the possibility to include additional everyday
objects to be a part of the proposed system in the future.
Generic communication boards are designed with easily
replaceable sensor connectors to facilitate multiple sensors by
only replacing the onboard sensor and microcode. Maxstream
XBee 802.15.4 transceiver and Atmel ATMEGA88-20PU
microcontroller are used in individual generic communication
boards. The XBee transceiver operates at ISM 2.4 GHz
frequency, 1mW (0 dBm) power output and allows for data
rates of up to 250 Kbps. The average data rate of all the
sensor nodes was 20.4 Hz (with a maximum of 100 Hz for
some nodes and a minimum of 10 Hz for a majority of the
nodes). The microcontroller is run at 8 MHz.
Sensing Precision and Recall values
The accuracy in sensing the object state changes based on
the user’s interaction with those objects is an important factor
to evaluate. The sensing system was evaluated within
scenarios where in the subjects were performing a set of
everyday activities. Such an approach was considered due to
the following evaluation criteria: 1) the sensing system should
be evaluated as a whole instead of sum of the individual parts
in isolation; 2) the sensing system should be evaluated in a
realistic setup where the subjects are performing everyday
activities by interacting with everyday objects; and 3) to use
the data collected for activity recognition with the additional
information known about the accuracy of the sensing system.
In our previous work [2, 16], we have used multiple Hidden
Markov Models (HMMs) in parallel to accommodate
individual information channels like objects
grabbed/released, objects’ state changes, and objects around
the user’s body for recognizing activities.
CONCLUSIONS
This paper has shown the results obtained from an actual
deployment of a WSN in a living laboratory home
environment for ubiquitous computing research. The state
changes to everyday objects present in the user’s environment
were sensed with an overall precision value of 91.2% and an
overall recall value of 98.8%. The signal strength measures
were also shown to be acceptable 97.5% of the time evaluated
in 8 different locations in the environment. The transmission-
reception range evaluated with wall obstructions common in a
home environment has also shown promise. Finally we have
also addressed some of the non-functional requirements
imposed by the subjects.