Abstract
After conducting some experiment one always faces the tedious work of manually
annotating and examining it. This is time consuming and error prone, especially
if the experiment contains a large amount of relevant events that need to be
annotated. In this thesis we tackle this problem for a real world experiment
from the field of sport psychology, where two players have to pass a ball to
each other as quickly as possible by bouncing it over a target zone. The
experiments evaluation consists of the number of passes that bounced on the
target zone in relation to the number of bounces that did not. Our main idea
was to cover the target zone with a surface different from the one outside of
it and use audio classification techniques to classify the bounces. To obtain
the required audio data we made minor changes to the experiments setup by
adding two microphones to it and
changing the surface of the target zone. We conducted a number of experiments
with this setup, extracted and analyzed all bounces from the audio data we
retrieved and showed, that an automatic classification is very possible and
feasible.