Georgia Southern University
Optimization of Filter Bank Canonical Correlation Analysis for Ssvep Based Brain Computer Interface
Abstract
dc:description.abstractSinusoidal steady state visually evoked potential signals, or SSVEP signals, are a type of brain signal generated through the user’s prolonged exposure to flickering visual stimuli. The canonical correlation analysis(CCA) method is a means of classifying this type of signal by comparing the frequency components of the recorded SSVEP signal to a generated reference signal. This method can be further improved through the use of a filter bank to divide the signal into multiple sub bands before performing analysis and combining the results to determine the frequency classification. The experiment performed in this paper tested five overall parameters to determine which settings would optimize performance for four frequency classifications at 8 Hz, 9 Hz, 10 Hz, and 11 Hz. The parameters tested included: number of reference harmonics, weights a and b, number of sub bands in the filter bank, and the stopband frequencies of the filter bank designs. The highest overall classification accuracy of 98.69 percent was found under four reference harmonics in a filter bank comprised of equally spaced start and stop bands, with four filtered sub bands, and a weight combination of �� = 2, �� = 0.75.
Degree
thesis:*- Name thesis:degree_name
- Master of Science, Electrical Engineering
- Level thesis:degree_level
- Thesis (restricted to Georgia Southern)
- Discipline thesis:degree_discipline
- Department of Electrical and Computer Engineering
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ferrara, Nicholas
- Contributors dc:contributor
-
- Dr. Mohammad Ahad
- Dr. Fernando Rios
Subjects
dc:subject × 8Rights
- Licence dc:rights.license
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.georgiasouthern.edu/etd/2784