Abstract:
This dissertation work deals with the design and development of a fuzzy controller to analyze electroencephalogram (EEG) data. The fuzzy controller made use of the multiple functions associated with the different regions of the brain to correlate multiple Brodmann areas to multiple outputs. This controller was designed to adapt to any data imported into it. The current framework implemented supports a math study and a police officer study. The rules for the interactions of the Brodmann areas have been set up for these applications, detailing how well the police subjects’ brains exhibited behavior indicative to activation relating to vision, memory, shape/distance, hearing/sound, and theory of mind. The math subjects’ outputs were attuned to their related study which involved transcranial direct current stimulation (tDCS), which is a form of neurostimulation. Anode affinity, cathode affinity, calculation, memory, and decision making were the outputs focused on for the math study. This task is best suited to a fuzzy controller since interactions between Brodmann areas can be analyzed and the contributions of each area accounted for.The goal of the controller was to determine long-term behavior of the subjects with repeated sampling. With each addition of data, the controller was able to develop new bounds related to the current condition of the data in the study. Processing this data was accomplished by the creation of an automated filtering script for EEGLAB in MATLAB. The script was designed to rapidly load and filter the files associated with any given dataset. These files were also automatically prepared for analysis with a program called Low Resolution Brain Electromagnetic Tomography i.e. (LORETA). LORETA was used to solve the inverse problem, which involves identifying where the signals from the surface electrodes originated within the brain through a process called source localization. Once the sources of the EEG signals were located, they were associated with the Brodmann areas. The fuzzy controller then processed this information to automatically generate heat maps which displayed information such as normalized data, z-score, and rankings. Each set of scores displays how the subject's brain was acting, which lined up with the expected results.
