Sliding Window and Filterbank Utilization on Riemannian Geometry


16th International Conference on INnovations in Intelligent SysTems and Applications, INISTA 2022, Biarritz, France, 8 - 12 August 2022 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/inista55318.2022.9894208
  • City: Biarritz
  • Country: France
  • Keywords: Brain-Computer Interface, EEG, Riemannian Geometry, Tangent Space Mapping
  • Kayseri University Affiliated: No


© 2022 IEEE.Riemannian geometry-based signal processing approaches on EEG signals provides similar decoding performance compared to state-of-the-art methods. However, Riemannian geometry framework requires predefine EEG signal epoch that is to be used in the analysis. Sliding window approach that operates in Riemannian geometry proposed to enable use of EEG signals without constrained by the record length. Decoding performance of tangent space mapping was increased more than 6% in overall accuracy compared the previous study's results. Instead of using single band-pass filter, utilization of filterbank is proposed to increase decoding performance. Distance based Riemannian classifier's overall performance were increased by 5% compared to standard Riemannian geometry approach.