Master thesis presentation: Brain-Computer Interfaces in a car: Science fiction or a realistic concept?
Presented by Martin Kaalhus, MPBME, and Adrian Edgren, MPSYS
Date: Friday, June 5th
Place: Lunnerummet (room 3311), Hörsalsvägen 11, 3rd floor
Examiner: Tomas McKelvey
This study aims to evaluate the feasibility of using a BCI (Brain-Computer Interface) base upon the P300 ERP (Event-related Potential) in an user-friendly setting in a vehicle. Compared to earlier studies within the field this thesis evaluates the use of BCI systems in a new setting, taking steps towards a commercial product. The specific end-application in mind is a strategic controller in an autonomous car, using the system to issue commands such as overtake or changing lanes.
Using an Emotiv EPOC headset a number of tests have been performed to identify the optimal set of components to use in the specific setting with the selected hardware. From the analysis of the data it can clearly be seen that the P300 is identified. Different signal processing and machine learning tools were applied and through evaluation of the performance it could be seen that using ICA (Independent Component Analysis) as a spatial filter combined with a SVM classifier for a 3×3 matrix of choices was the ideal choice of components for the current setup.
Combining these parameters the system gives strong indications of issuing commands with accuracy and repeatability in a car setting, with all inherent disturbances such as glare, vibrations, sound etc. This shows great potential to develop a functional BCI for implementation as a strategic controller, with further development and revision of the used hardware. A major part in further development of a system considers the usability and repeatability, since the electrode positions should be the same for each use and re-calibration as well as initiation time for the user should be kept at a minimum. One important example is that the current electrodes