Description

1. Protocol: Seven male and three female subjects (age 25 to 30), who have experienced aggression in scenarios such as physical fighting, took part in the experiment. Throughout 20 individual experiments, each subject had to perform ten normal and ten aggressive activities. Regarding the rights of the subjects involved, ethical regulations have been followed based on the code of ethics of the British psychological society, which explains the ethical legislations to conduct statistical experiments using human subjects. For safety precaution issues, boxing hand wraps have been given to the subjects, and for the warm up the subjects were instructed to familiarise with the bag by having a number of trial runs. The subjects were aware that since their involvement in this series of experiments was voluntary, it was made clear that they could withdraw at any time from the study. 2. Instrumentation: The Essex robotic arena was the main experimental hall where the data collection took place. With area 4x5.5m, the ten subjects expressed normal and aggressive physical activities at random locations. For the normal actions, a human partner has been used as a focus target attracting the attention from the subjects so as to perform more realistic activity. For the aggressive actions, the subjects made use of a professional kick-boxing standing bag, 1.75m tall, with a human figure drawn on its body. The bag has cylindrical shape made from soft material, which could bounce when hit. All the activities have been recorded from random starting positions so that to have a variety of spatial 3D data. The subjects performance has been recorded by the Vicons nine ubiquitous cameras, interfacing human activity with spatial coordinate points. Based on this context, the data acquisition process involved four reflectable markers placed on the forearms (elbows and wrists), four on the forelegs (knees and ankles), and one on the top of the head. 3. Data Setup: Each experimental trial has been taken separately for each physical activity. The duration of each action was approximately ~10 seconds per subject, which corresponds to a time series of ~3000 samples, with sampling frequency of 200Hz. Within this performance time, approximately 15 action trajectories were extracted counting in average 15 normal (ex: handshaking), and 15 aggressive (ex: punching) actions.