KEYWORDS:
ABSTRACT:
The project "Human Motion Processing in Sequences"
is based on a general investigation into methods and algorithms of computer
vision for static pictures and spatio-temporal image data. This work has
provided the background for the creation of an enviroment for future investigations
and implementations of the new information processing group at the Knowledge
based Systems section. The application that has resulted from this work
constitutes the first attempt to use the spatio-temporal data for an analysis
of human emotion, and, to derive expressive movements patterns from a video
input signal. The procedure of relating emotional states of human observant
to the input signals reposes on a parametric description of the visual
and psychological events and states. The underlying models were constructed
from the different fields of this interdisciplinary project. Their combination
and the adaptation of the associated flow of data to a given structure
of implementation resulted in an engineering task that will be described
in this thesis.