Abstract
In this dissertation, I focused on EEG classification for high-level cognitive activities. I first reviewed the machine learning and deep learning algorithms that already have been implemented in EEG classification and summarized the categories of the state-of-the-art algorithms, which set up the baseline for my experiments.
Then I started collecting data, mainly instructing non-expert end-users to collect their EEG while performing multiple tasks in home-like settings with non-invasive, consumer-grade EEG devices. To gather more data faster and more efficiently, I introduced a new framework, EEG4Home. This framework could make the EEG data collection and analysis process smoother for large-scale data collection from non-expert participants at home. Such a framework could be helpful for future machine learning and deep learning algorithms development and its implementation in EEG-based BCI.
Then I proposed a new algorithm, Time Continuity Voting (TCV), experimentally evaluated it compared to a dozen state-of-the-art algorithms, and demonstrated its advantage of achieving high performance with reasonable computational time across four experiments and one hundred and sixteen healthy participants.
I concluded that this TCV algorithm has the potential to be a promising algorithm for EEG classifications and discussed the significance and broader impact.