An Artificial Perceptual Model Approach - Artificial Human Sensors: Science and Applications - page 125

observation-based experiments. These experimentscanperformasajointsystem-
operator verification, and can be a useful approach in performing tools for verify-
ing the system's performance.
In artificial intelligence (AI), a verification methodology is proposed in Cohen
(1995), where experimental results are data, often presented as quantitative repre-
sentations of the system behaviour, that are further processed under consideration
and tried to estimate the performance. The numerical data can then be sorted in
three qualitatively related types depending on how the parameters relate to each
other:
Categorial-, Ordinal- and Numerical data.
In case of artificial perceptual systems, a direction towards an adjustment is
needed. Additional sets of data have to be specifically performed in proposing the
following distinct types:
Categorial data: is regarded as sets of classes. For instance, the senses may
record a data set as classes in vision, auditory, olfaction
during a time period.
Ordinal data: can be put in order or ranked as a priority. For example, in
an experiment the vision sense has provided more infor-
mation than auditory and olfaction, which have a minor
effect on the experimental result.
Numerical data: comprises quantitative data, for example, the numerical-
values of hue, saturation or temperature of colours.
Further the three qualitative and distinct types of data classification can be ex-
tended with the following processing types:
Fusional information: is related to the fusing algorithm, that evaluate and esti-
mate the information gained from the received data type.
For example, taste lasts longer than smell in aged people.
Interface status: will be observatory experiment, indicating the impression
of the added information in strengthening or weakening
the individual decision-making. For example, I am able to
notice that the surrounding light is very dim for this type
of equipment.
Obviously, we have to take into account a traditional statistical experimental
methodology for the artificial perceptual system and an observatory methodology
for the human part of the overall system. The performance as shown in Fig. 5.16 is
actually shown in the individual's satisfaction and the added value of the increased
human capability.
In a confirmative aspect of complex experiments to verify performance of sen-
sor systems, there is a need to observe obtained effects in the output. There are
various statistical tools to calculate the requested relationships and to visualise the
resulting output, as for example was mentioned in the previous chapter.