The public defense of Shaibal Barua's doctoral thesis in Computer Science and Engineering
The public defense of Shaibal Barua's doctoral thesis in Computer Science and Engineering will take place at Mälardalen University, room Zeta, MDH Västerås, at 13.15 on February 21, 2019.
Title: “Multivariate Data Analytics to Identify Driver’s Sleepiness, Cognitive load, and Stress”.
Serial number: 284
The faculty examiner is Professor Nirmalie Wiratunga, Robert Gordon University, and the examining committee consists of Associate Professor Arne Lowden, Stockholm University; Associate Professor Maria Riveiro, University of Skövde; Associate Professor Sławomir
Nowaczyk, Halmstad University.
Reserve: Associate Professor Saad Mubeen, Mälardalen University
Road transportation is a complex system and poses challenges in safe transportation due to dynamic environment changes and adaptation of driving according to these changes. Impaired driving is dangerous and factors such as sleepiness, cognitive load, stress often causes car crashes. In literature more than 90% of traffic crashes are assigned to the drivers. Futuristic driving paradigm focuses on detection of the driver states for current naturalistic driving and also for a highly automated driving system. To achieve that goal the driver should be kept in the loop and it requires to understand and identify the objective measures. Hence, domain knowledge from sleepiness, cognitive load and stress can be used with machine learning for detection and classification of these driving states. For detecting and classification of sleepiness, cognitive load and stress the aim is to develop an objective measure of sleepiness, cognitive load, and stress that can be used as a research tool, either to benchmark unobtrusive sensor solutions, or when investigating the influence of other factors on sleepiness, cognitive load, and stress.
This thesis investigates the use of standard sources of information to detect sleepiness, cognitive load, and stress. It also explores the improvement of detection and classification accuracies by adding contextual information and data fusion. In this thesis multivariate data are used for classifications. This thesis work includes several physiological signals namely, electroencephalography (EEG), electrooculography (EOG), electrocardiography (ECG), skin conductance, finger temperature, and respiration signal. Among the multivariate signals, electroencephalography (EEG) signal is thoroughly investigated for pre-processing step i.e., noise and artifacts handling. In addition, contextual information and driver behavioural data are analysed and used along with physiological data for feature engineering. Several machinelearning algorithms along with different data and signal processing techniques have been investigated for classification and detection of sleepiness, cognitive load, and stress.