Loading...

Improving CCA Based Methods for SSVEP Classification using Graph Signal Processing

Noori, Nastaran | 2023

68 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56647 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Hajipour Sardouie, Sepideh; Einizadeh, Aref
  7. Abstract:
  8. The Brain Computer Interface (BCI) translates brain signals into a series of commands, enabling individuals to fulfill many of their basic needs without physical activity. Electroencephalogram (EEG) signals are commonly used as input for BCI systems, because the recording of this signal is non-invasive, inexpensive, and also have an acceptable time resolution. One of the most prevalent methods in BCI systems is the brain-computer interface based on Steady State Visual Evoked Potentials (SSVEP). These systems provide high response speed and Information Transfer Rate (ITR) as well as a good signal-to-noise ratio (SNR). The main purpose of these systems is to detect the frequency of SSVEP in order to understand which visual stimulus a person is looking at. This information can be converted into a command to assist individuals with disabilities. Various methods have been proposed to detect the frequency of SSVEP. One of the most widely used methods is the Canonical Correlation Analysis algorithm (CCA). This algorithm has been widely used due to its simplicity in implementation and high accuracy, and is the basis of many studies. The aim of this thesis is to improve CCA-based methods for detecting SSVEP frequency using Graph Signal Processing (GSP) methods. To this end, the graph structure present on the time samples of the EEG signal is used. Using GSP methods, we improved the CCA method in two separate parts. In the first part, methods to improve the CCA cost function by adding graph constraints were proposed. As a result, the GCCA, FoGCCA, IT-GCCA and finally IT-FoGCCA methods were obtained, with the IT-FoGCCA method achieving the highest accuracy of 83% in a 1.5-second time window. In the second part, methods for creating a new reference signal in the CCA algorithm were presented. In this section, the existing graph among the time samples was used, and based on this, three algorithms GL-ITCCA, WGL-ITCCA and HGL-CCA were introduced, with the best result obtained for the WGL-ITCCA algorithm
  9. Keywords:
  10. Electroencphalogram Signal ; Graph Signal Processing ; Brain-Computer Interface (BCI) ; Steady State Visual Evoked Potential (SSVEP) ; Canonical Correlation Analysis ; Frequency Recognition

 Digital Object List

 Bookmark

No TOC