Identification of Affect Patterns from Bio-signals

Abstract

In her seminal work, Rosalind Picard defined Affective Computing as, computing that relates to, arises from, or deliberately influences emotions. The aim of the thesis is to explore the possibility of application of affective computing concepts to the DLR Robotic Motion Simulator(DLR-RMS). The DLR-RMS is a serial kinematics based platform that uses an industrial robot to provide the motion cues to an attached simulation cell. By making the DLR-RMS emotion aware, the simulator can e.g., adapt simulations to the subjective needs of its users. This thesis is one of the first steps taken towards fulfilling the aforementioned goal and therefore, concerns itself primarily with laying the ground work for further studies. To do the same, a ground based (not on the simulator) experiment that uses video clips as stimuli (instead of video and motion, as on the simulator) was designed and undertaken to acquire the emotional/affective response from 30 human subjects. A novel user feedback interface, which employs a joystick to acquire subjective evaluation for video clips, was also developed. This interface allows the subject to continuously state his/her emotional status that was elicited by viewing the video clips. The data acquired during the experiments was analysed using statistical techniques e.g., Principal Component Analysis (PCA) to determine the trend in the recorded data. The thesis concludes by presenting the results of the data analysis, which show favourable and expected patterns in data that validate the design and methodology of the experiment and lay the ground work for further experiments to be undertaken on the DLR-RMS.