Civil Engineering Reference
In-Depth Information
Fig. 6
Introduction to the concept of frame, state, scene and spectrum
Given this description, the ECM is a cluster of n different MC, where n rep-
resents the number of spaces in the building, as represented in Fig.
4
. Accordingly,
the ECM can be written as
ECM
¼
[
n
MC
i
ð
k
Þ
;
8
MC
i
ðÞ¼
X
i
k
þ ð Þ¼
AX
ð
k
Þ
Y
i
ðÞ¼
CX
i
ð
k
Þ
ð
4
Þ
i
¼
1
The next step consists of estimating the ECM parameters based on acquired
measurements. That is, estimate the probability transition matrix P for each MC
i
of
the ECM. To solve this problem, an algorithm is proposed named Frame-Scene
Analysis.
3.3 Frame-Scene Analysis
A strong assumption for the MC is that system states being modelled are obser-
vable. So in order to obtain the model parameters, it is fundamental to have access
to the spaces' states, since they must be observable. However, the measurements
are a collection of visible events and not directly the states. To estimate the MC
parameters, it is necessary to reconstruct the spaces' states based on the available
events.
The developed algorithm was inspired by the work of Wada and Matsuyama
(
2000
) regarding behaviour recognition in a video sequence. The concept is that
during the observation period the spaces suffer changes to their configuration due
to the presence of the occupants and their actions. As a result, the space's states
evolution is analogous to a video sequence in which there exists a history evolving
over time.
