Brain-computer interfaces (BCIs) provide means for direct braincomputer interaction, based solely on the user's brain neural activity, commonly captured by Electroencephalography (EEG), and do not rely on any degree of physical movement. From a general perspective the function of BCIs is to discriminate between a limited set of mental states, which the user enters voluntarily or unconsciously. This represents a foundation for various BCI applications such as assistive technologies, including neuroprosthetics and computer control BCIs for disabled users or mental state monitoring systems aimed for emotion, fatigue or workload recognition. A commonly used type of mental tasks for BCI control is imagination of physical movement or motor imagery, which is characterized by the local power deviation occurring in the brain areas responsible for muscles involved in the executed task. This PhD manuscript is dedicated to the design of motor imagery EEG BCIs with a particular focus on signal processing and classification approaches that incorporate the background knowledge about biophysics and EEG signal generation. These aspects are considered in the EEG source reconstruction process, which estimate the cortical currents during the EEG voltage measurements from head surface. In this work it is shown that the application of the source reconstruction in a BCI signal processing scheme effectively decreases the negative effects of EEG electrode coupling providing for an increase in class separability, given that the cortical areas involved in motor imagery are anatomically segregated. Based on these observations a novel BCI feature extraction method based on source analysis and common spatial patterns (CSP) was proposed and its performance was investigated with a common motor imagery dataset and our own real-time BCI implementation. Our results show that EEG source reconstruction reduces the influence of noise and muscular artifacts, and thus the proposed approach consistently outperforms the conventional BCI sensor feature extraction methods.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:713313 |
| Date | January 2017 |
| Creators | Zaitcev, Aleksandr |
| Contributors | Cook, Greg ; Liu, Wei ; Paley, Martin ; Milne, Elizabeth |
| Publisher | University of Sheffield |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://etheses.whiterose.ac.uk/17155/ |
Page generated in 0.1977 seconds