Toward Brain-Computer Interfaces in Real Life: Solving the Interferences to the Use of Event-Related Potential-Based BCIs

Abstract

A Brain-Computer Interface (BCI) is a system that interprets neural signals to control external devices or facilitate communication, offering significant benefits by bypassing muscle pathways. Despite extensive research, practical applications in daily environments remain limited due to several challenges. Especially, recent advancements in augmented reality (AR) and virtual reality (VR) technologies have broadened BCI applications but also introduced new challenges, such as increased cognitive load due to external interferences. This dissertation investigates the performance of event-related potential (ERP)-based BCIs in realistic settings, addressing challenges posed by external interferences and exploring solutions to enhance stability and usability. The research is structured into three main chapters. The first chapter involves the development and evaluation of an AR integrated online ERP-based BCI platform aimed at controlling home appliances. This chapter elaborates on the integration of see-through user interface (UI) with augmented reality to enhance the user's interaction with the system in a real-world setting. The BCI system, connected through internet of things (IoT) and controlled via AR interfaces, was tested for its efficiency in manipulating common household appliances like televisions, door locks, and lights. The ERP-based BCI system implemented in this study served as the foundational platform for subsequent research endeavors. The second chapter examined the effects of external interferences on the performance of ERP-based BCI systems, specifically categorizing them into distractions and interruptions. The findings from the two studies presented in this chapter underscore the significant impact of different types of interferences on the accuracy and reliability of BCI operations. The first study aimed to investigate the effects of auditory distractions with emotional conditions on ERP-based BCI performance. However, it was found that the influence of auditory distractions, including emotionally charged auditory stimuli, was not significant in modulating ERP amplitudes crucial for effective BCI control. The second study explored the impact of active verbal communication tasks on ERP-based BCI performance. It was observed that real-life tasks involving speaking act as substantial interruptions, requiring the user to reallocate cognitive resources away from the BCI task. These interruptions lead to a noticeable decrease in the ERP components' amplitudes including P3 and N2, which are essential for accurate BCI operations. In contrast, tasks like active listening and simple syllable pronunciation showed no substantial impact on BCI usability. Additionally, the study identified a correlation between occipital N2 and alpha power with BCI accuracy, indicating that speaking particularly diminishes visual attention, thereby adversely affecting BCI performance. The third chapter explored the implementation of an adaptive model in ERP-based BCIs to enhance performance in interrupting environments. Two primary studies were conducted to evaluate the effectiveness of the adaptive model under different types of interruptions. The first study focused on the application of a multi-window adaptive approach to manage interruptions effectively. It was found that employing multiple time windows for analysis allowed the system to maintain higher classification accuracy even when users were subjected to speaking interruptions. This finding confirms the potential effectiveness of using this model in scenarios where interruptions are present. The second study investigated the performance of the multi-window adaptive BCI model during video-watching tasks, a common real-world scenario that involves significant visual interruptions. The results demonstrated that interruptions caused by video-watching decreased ERP amplitudes and consequently reduced BCI accuracy. However, employing the multi-window adaptive model significantly outperformed non- adaptive ones, maintaining accuracy and partially restoring performance levels despite interruptions. Overall, this dissertation makes several key contributions to the field of BCI research. It provides a comprehensive analysis of how different types of external interferences affect ERP-based BCI performance, validates the effectiveness of adaptive models in mitigating these effects. The research represents a crucial step towards the broader adoption of BCIs in everyday life, demonstrating that with appropriate adaptations, BCIs can be made robust enough to function effectively in a wide range of real-world settings with various cognitive interferences.