Brain-controlled virtual environments : an evaluation study of Brain computer interfaces for serious game interaction

摘要

Brain-Computer Interfaces (BCIs) have been increasingly used the last few years in research as a control input to various BCI games, incorporating different techniques and methodologies. So far all current brain-controlled games and virtual environments are proof-of-concepts, with low information-transfer-rate, lack of 3D graphics or poor game-play resulting to poor immersion. Within the last few years commercial BCI equipment had been introduced as an alternative game controller from various companies like Emotiv and Neurosky, setting a milestone in user experience. Although new games are emerging which make use of these (out-of-the-box) headsets, there is not much knowledge concerning the game design principles that have to be taken into account to create a game or virtual environment that is controlled by a unique device like EEG headsets bearing in mind that even neuroscientists have just started (with modern methodologies and tools) to map the human brain and it’s functions. This study presents the background in BCI’s and a state-of-the-art from current BCI controlled serious games and Virtual Environments (VEs), investigates the effect on the user experience this kind of brain-to-computer interface has. In addition, it examines whether or not the incorporation of a BCI in Serious Games and VEs can offer similar control and engagement that other Natural User Interfaces (NUIs) can offer. To achieve this, a pilot study of n=31 naïve participants (with no previous experience with BCI’s) had been conducted, gathering feedback through a presence questionnaire, an unstructured interview and Encephalographic (EEG) data during task performance. For the evaluation process, two serious games were used. The first was designed for controlling a robot in a virtual maze (BrainMaze) in a third person perspective and the second game was “Roma Nova”, a high-fidelity serious game created in Coventry University. Results indicate that users are capable of controlling a virtual avatar followed a training task. User self-report questionnaires indicate enjoyment and acceptance of the proposed system. Finally, the correlation between the user answers based on the gaming experience and the extracted EEG data for engagement and attention is very small indicating that current BCI technology is not capable to deliver the same qualities of interaction as other Natural User Interaction (NUI) devices that are currently on the market. Furthermore, this study sets the base for the development of future BCI driven serious games. The objectives of this thesis are threefold. First, to control an avatar of a serious game in real-time performance using only EEG data; secondly, to examine the reaction of users while playing the game and thirdly to test each system in terms of: ‘learnability of the interface using the game’, ‘satisfaction of the player’, ‘performance of the interfaces’ and ‘effort put by the player’. This will help us to build a more generic framework for future brain-controlled games.