Dissertation/Thesis Abstract

EEG-Korrelate der Fehlerverarbeitung während der Fahraufgabe
by Protzak, Janna, Ph.D., Technische Universitaet Berlin (Germany), 2015, 183; 10717097
Abstract (Summary)

Despite ongoing improvements to vehicle safety and development of advanced driver assistance systems, safety critical situations in road traffic occur. There is still no reliable method to assess and evaluate individual driving performance and driver errors, and which, through adaptive assistance systems, can further support the driver in potentially dangerous situations. The analysis of brain waves, measured through electroencephalography (EEG), could be utilized to identify and analyze parameters that provide useful information on human perception and error processing. The main objective of this thesis is to evaluate applications of EEG- measurements under realistic driving conditions on the basis of selected, exemplary scenarios of error evaluation processes. For two test series, data were collected under controlled laboratory conditions and in subsequent, real driving maneuvers. Data analysis comprises of basic, research-orientated analysis of event-related activity as well as the use of application-orientated classification algorithms from the field of Brain Computer Interface (BCI) research. The first experimental series addresses neuronal processes underlying erroneous accelerator and breaking pedal applications during a simple stimulus response task. The data collected under laboratory conditions revealed the expected differences in the averaged neuronal activity between correct and incorrect pedal responses. Incorrect pedal applications are followed by event related error potentials that are composed of Error Related Negativity (ERN) and a subsequent Error Positivity (Pe). Classifications performed on the data revealed 90% accuracy. However, this compelling result is attenuated by a high rate of false alarms. Error potentials appeared only slightly in the realistic driving maneuver data and are accompanied by a second neurophysiological correlate, known as the Contingent Negativity Variation (CNV). Nonetheless, the results indicate clearly that a meaningful data analysis of the data collected from an actual driving task is possible–despite the numerous artifacts and noise sources present in a realistic vehicle environment. Classification accuracy for these data sets reached 74%, although error detection was only slightly better than chance (57 % error detection rate). The second approach focused on errors that may occur during the interaction with a technical system. A memory task was used to simulate the interaction between participants and a speech based input system that was randomly interrupted by artificially produced erroneous feedback. Feedback that didn’t correspond to the users previous input elicited a P2- P300-Slow Wave Complex in the averaged EEG-Data. Meanwhile correct answers only evoked a P2-Potential. The overall classification accuracy for these data resulted in a high rate of 84% correct classifications. Nevertheless, again, high rates of false alarms must be taken into account. Therefore, the results obtained from the data of the laboratory studies were replicated by the data analysis of the realistic driving maneuver: The classification rate of the Driving task data was 74%. However, as in the previous study, errors were detected in only 57% of all cases. The results of these studies show that parameter of complex cognitive processes, like error processing and the evaluation of information in real-world driving situations, can be detected via EEG. In particular, the findings of the experimental series on interaction errors demonstrate clearly, that the results of the actual driving task are comparable to results determined under controlled laboratory conditions and, thus, are suitable for scientific analysis, despite the artifact bonded vehicle environment. These findings further illustrate the potential of EEG-technology to obtain information about cognitive processes that could contribute to the design of innovative passive-adaptive assistance systems.

Indexing (document details)
Advisor: Manzey, Dietrich
Commitee: Jürgensohn, Thomas, Rötting, Matthias
School: Technische Universitaet Berlin (Germany)
School Location: Germany
Source: DAI-C 81/1(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Transportation, Public health
Keywords: EEG, Error processing, Driving
Publication Number: 10717097
ISBN: 9781392605516
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest