| |
5.2. The Sensor-Opportunities-Based Approach
65
Physical View: Placement of Sensors
An application engineer has to consider the possibilities of sensor placement in the
physical world: e.g. a tra c jam can be remotely detected by a camera or locally at
each car through mutually exchanged distance and speed information. Both choices
are appropriate for the intended purpose but have di erent side-e ects: the camera
has to be mounted once and works for all cars, but only at one specific location. As a
side-e ect its use could be extended to other applications, e.g. criminal search. The
local set-up instead requires individual e ort at each car but therefore users have
the control participating in the system or not. And the system works everywhere.
We identify four di erent categories of sensor placement. In Environment refers
to stationary installed sensors, e.g. in the floor, walls, where placement can only
be changed with e ort. Whereas In Environment installations work with all users
at the stationary location On Human has the opposite characteristics: only users
wearing the sensors can participate, but therefore they are not bound to a location.
On Object is in between the two previous categories, as objects can be personal and
can be carried with a human (e.g. key), but also stay at a certain location (e.g.
chair). This distinction depends on the object. Additionally, mutual Collaboration
defines sensing system that always require more than one unit in order to operate
properly, e.g. triangulation of signal strength for localization.
Review: Sensors in Ubiquitous Computing Research
Based on own experience with sensors and literature review we compiled a table
(Figure 5.1) characterizing sensor technology with respect to the six sensing di-
mensions and the four sensor placement possibilities. This table should be used as
reference for application developers during the process of finding the appropriate
sensors for their application.
In each cell of the table sensors are aligned due to bandwidth consumption and
quality of perception in respect to the dimension. The alignment due to precision
and bandwidth should be seen as a rough estimation for relative comparison between
sensors occurring in the same table field. It follows a line-wise discussions of the
cells of the Table 5.1.
For recognizing a person’s ID the table shows four choices of sensors for installation
In Environment in the upper left cell. Obviously, the best results can be achieved
with biometric sensors
[
Wayman et al. 2003
]
, such as finger print or iris scan, as
represented by vertical alignment in the cell. Methods based on vision
[
Donato
et al. 1999
]
, audio or load-cells embedded into the floor
[
Cattin 2001
]
deliver less
quality. Horizontal alignment in the cell shows, that data generated by load-cells
and finger print sensors consumes lower bandwidth than methods based on vision or
|  |
|
| |
|
|
