Abstract

Recent advancements in Electroencephalography (EEG) sensor technologies, Signal Processing (SP), and Machine Learning (ML) algorithms have paved the way for further evolution of Brain Computer Interfaces (BCI) in several practical applications ranging from rehabilitation systems to smart consumer technologies. In particular, this thesis research is motivated by potentials of BCI platforms to provide comfortable means for individuals with communication disabilities to interact with the outer world. When it comes to SP/ML models for BCI systems, there has been a surge of interest on Visual Evoked Potentials (VEPs). Recently, Steady-state visual evoked potential (SSVEP) has become popular due to their fast and reliable performance, and strong provocation of visual brain signals. Despite the popularity of SSVEPs, their utilization for practical applications especially for assistive technologies is challenging due to eye fatigue and risk of induced epileptic seizure. In this regard, the key issue of conventional light-flashing techniques has been addressed by development of flicker-free Steady-State motion Visual Evoked Potential (SSmVEP). Such benefits, however, come with the price of having less accuracy and less Information Transfer Rate (ITR). In this regard, the thesis focuses on improving the following three main components: (i) Stimulation paradigm; (ii) Frequency modulation, and; (iii) Target classification in SSmVEP-based BCIs. With regard to the first component, novel SSmVEP paradigms with low luminance contrast and oscillating expansion and contraction motions are designed, and integrated within a BCI system. Through experimental evaluations, high detection accuracies are achieved for our proposed paradigms leading to less visual tiredness in comparison to conventional SSVEPs. Concerning the second component, an efficient modulation mechanism is proposed without using resources such as trial time, phase, and/or number of targets to enhance the ITR. The proposed design is based on the intuitively pleasing idea of integrating more than one motion within a single SSmVEP target stimuli, simultaneously. To elicit SSmVEP, we designed a novel and innovative dual frequency aggregated modulation paradigm, referred to as the Dual Frequency Aggregated steady-state motion Visual Evoked Potential (DF-SSmVEP). The proposed DF-SSmVEP is evaluated based on a real EEG dataset and the results corroborate its superiority. With respect to the third component, it is expected that incorporation of human brain's nonlinear dynamics and characteristics of the designed videos within our EEG signal classifier lead to a comprehensive model resulting in better noise removal. To this end, a deep learning-based classification model is proposed, referred to as the Deep Video Canonical Correlation Analysis (DvCCA), that extracts features of the SSmVEPs directly from the videos of stimuli. The proposed DvCCA is evaluated based on a real EEG dataset and the results corroborate its superiority against recently proposed state-of-the-art Convolutional Neural Network-based models.