Sensor-Based Knowledge Acquisition and Signal Processing Services - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 264

Part III
Sensor-Based Knowledge Acquisition and
Signal Processing Services

Overview

Traditionally, most information that was recorded in history was conveyed in
textual form. Think, for example, of the works of great philosophers, novelists, or
historians who relied on the written word to express their view points, to express
their creativity, or to report what they witnessed. Without any doubt, information in
human-readable textual form is the most commonly used means of sharing infor-
mation. At the same time, however, information is often shared in non-textual form.
Early art work such as cave paintings dating back over 30,000 years indicate that
mankind has always relied on many other forms to convey information. With the
invention of photography and photographic

lmmaking, the share of non-textual
documents that was used to distribute information increased signi cantly. In the

early days of the twentieth century, when photography started to take off, the adage
“ a picture is worth a thousand words ” was used to express the notion that complex
ideas can easily be conveyed in a still visual image. From a computational point of
view, interpreting non-textual material such as images or

lms is a nontrivial task.
The main challenge is to bridge the so-called semantic gap, i.e., the difference
between humans interpretation of the information that is depicted in such material,

and the representation of this information that can be processed by a machine. The
challenge of interpreting non-textual data even increases when considering non-
visual material such as sensor readings. With the introduction of sensors and sensor
systems, an ever-increasing amount of data is created that conveys detailed infor-
mation in the form of digital or optical signals. Analyzing this data for the provision
of information services requires advanced methods in the

'

elds of data mining and

machine learning. In the nal part of this topic, we present four different scenarios

where data in non-textual form is created and present approaches to exploit this
data.
The rst scenario, presented in Chap. 10 , addresses sustainable energy con-

sumption in smart home environments. More precisely, Spiegel presents a frame-
work for heating control and scheduling that considers occupancy information and,
therefore, allows for the reduction of residential energy consumption. His work
demonstrates how to use aggregated energy signals, measured by a smart meter, to