Cover
Vol. 15 No. 1 (2019)

Published: July 31, 2019

Pages: 13-27

Original Article

Session to Session Transfer Learning Method Using Independent Component Analysis with Regularized Common Spatial Patterns for EEG-MI Signals

Zaineb M. Alhakeem 1 Ramzy S. Ali 2
1Electrical Engineering Department University of Basrah Basrah-Iraq
2Electrical Engineering Department University of Basrah Basrah Iraq
History
  • Received: May 31, 2019
  • Available Online: July 31, 2019
DOI

https://doi.org/10.37917/ijeee.15.1.2

Abstract

Training the user in Brain-Computer Interface (BCI) systems based on brain signals that recorded using Electroencephalography Motor Imagery (EEG-MI) signal is a time-consuming process and causes tiredness to the trained subject, so transfer learning (subject to subject or session to session) is very useful methods of training that will decrease the number of recorded training trials for the target subject. To record the brain signals, channels or electrodes are used. Increasing channels could increase the classification accuracy but this solution costs a lot of money and there are no guarantees of high classification accuracy. This paper introduces a transfer learning method using only two channels and a few training trials for both feature extraction and classifier training. Our results show that the proposed method Independent Component Analysis with Regularized Common Spatial Pattern (ICA-RCSP) will produce about 70% accuracy for the session to session transfer learning using few training trails. When the proposed method used for transfer subject to subject the accuracy was lower than that for session to session but it still better than other methods.

Keywords

References

  1. H. Lu, K. N. Plataniotis and A. N. Venetsanopoulos, " Regularized Common Spatial Patterns with Generic Learning for EEG Signal Classification", Annual International Conference of the IEEE Engineering in Medicine and Biology Society , (2009).
  2. H. Ramoser, J. Müller-Gerking, and G. Pfurtscheller, " Optimal Spatial Filtering of Single-Trial EEG During Hand Movement", Transaction on Rehabilitation , 8(4), (2000).
  3. H. Lu, H. Eng, C. Guan, K. N. Plataniotis, and A. N. Venetsanopoulos, " Regularized Common Spatial Pattern With Aggregation for EEG Classification in Small-Sample Setting", IEEE Transaction on Rehabilitation , 57(12), 2010.
  4. R. Corralejo, R. Hornero, and D. Álvarez, " Feature Selection using a Genetic Algorithm in a Motor Imagery-based Brain-Computer 33rd Annual Conference of the IEEEE MBS Boston, Massachusetts USA, 2011.
  5. S. Soman and Jayadeva, "High-performance EEG signal classification using classifiability and the Twin SVM", Applied Soft Computing , pp. 305–318, 2015.
  6. Z. Tang, C. Li and S. Sun, " Single-trial EEG classification of motor imagery using deep convolutional neural networks", Optik, pp. 11–18, 2017.
  7. Y. Tabar and U. Halici, " A novel deep learning approach for classification of EEG motor imagery signals", Journal of Neural Engineering , pp. 11,2017.
  8. F. Lotte, L. Bougrain, A. Cichocki, M. Clerc, M. Congedo, A. Rakotomamonjy, and F. Yger, " A review of classification algorithms for the EEG-based brain-computer interfaces: a 10year update", Journal of Neural Engineering, pp. 28, 2018.
  9. http://www.bsp.brain.riken.jp/~qibin/homepage/Datasets.ht ml .
  10. Dataset IVa for the BCI Competition III.
  11. A. yva¨rinen, E. Oja:" Independent component analysis: algorithms and applications", Neural Networks , 13, 411– 430,2000.
  12. N.Alamdari, A. Haider, R. Arefin, A. K. Verma, K. Tavakolian and R. Fazel-Rezai, " A Review of Methods and Applications of Brain-Computer Interface Systems", IEEE , 2016.
  13. S.Sanei and J.A.Chambers, "EEG Signal Processing", Wiley , ISBN 978-0-470-02581-9, 2007.
  14. F. Lotte, M. Congedo, A. Lécuyer, F. Lamarche and B. Arnaldi, "A review of classification algorithms for the EEGbased brain-computer Journal of Neural Engineering , 4, pp.24, 2007. number of trials) Table II: Comparison of different feature extraction methods for (Dataset Ia, M=20, (session to session transfer learning)) Main session Power of alpha and beta CSP RCSP c_day1 49.1667 50.8333 c_day1_2 53.0769 51.5385 50.3077 65.3846 c_day1_3 38.7 c_day1_4 c_day1_5 58.8235 31.6176 83.8235 74.5588 c_day1_6 62.5 87.5 68.625 76.25 c_day1_7 71.5 85.1 Average 53.22387 56.49849 59.4927 68.0489 Standard deviation 4.964336 19.04735 16.20847 11.93954 Accuracy Number of trails Table III: Comparison of different feature extraction methods for (Dataset Ia, M=12, (session to session transfer learning) Main session Power of alpha and beta CSP RCSP c_day1 47.6563 51.5625 c_day1_2 54.3478 54.3478 51.5217 67.1739 c_day1_3 37.037 46.2963 58.33333 55.5556 c_day1_4 20.3704 45.3704 62.1296 c_day1_5 45.1389 58.3333 83.9583 82.1528 c_day1_6 62.5 85.2273 77.3864 83.1818 c_day1_7 53.7037 88.8889 69.9074 78.7037 Average 50.05481 57.86093 64.41292 71.4829 Standard deviation 7.425942 21.72283 13.82199 10.50858 1: Power of alpha and beta 2: CSP 3: RCSP 4:ICA_RCSP Accuracy Table IV: Comparison of different feature extraction methods for (Dataset Ib, M=12 (session to session transfer learning)) Main session Power of alpha and beta CSP RCSP c1 48.7421 88.0503 c1_2 58.3333 86.9048 76.7857 78.5714 c1_3 76.6667 73.3333 74.4444 Average 52.35847 83.87393 75.0595 76.5079 Standard deviation 4.255941 5.117695 1.7262 2.0635 1: Power of alpha and beta 2: CSP 3: RCSP 4:ICA_RCSP Accuracy Table V: Comparison of different feature extraction methods for (Dataset Ic, M=12 (subject to subject transfer learning)) Main subject Power of alpha and beta CSP RCSP A 50.5952 76.1905 80.5952 82.0238 B 59.6154 79.8077 72.0192 73.0769 C 48.1481 45.3704 50.5556 Average 52.78623 67.12287 67.72333 68.3669 Standard deviation 4.931208 15.45204 12.63422 13.49121 1: Power of alpha and beta 2: CSP 3: RCSP 4:ICA_RCSP Accuracy Table VI: Comparison of different feature extraction methods for (Dataset II, M=12 (subject to subject transfer learning)) Main subject Power of alpha and beta CSP RCSP aa 44.8718 46.1538 51.2821 50.641 al 62.2642 58.9623 50.9434 52.3585 av 39.1509 55.6604 50.9434 55.6604 aw 43.8679 46.2264 54.2453 57.0755 ay 42.4528 53.7736 Average 46.5215 52.1553 51.48284 53.14708 Standard deviation 8.10525 5.1460612 1.445841 2.776893 1: Power of alpha and beta 2: CSP 3: RCSP 4:ICA_RCSP Accuracy subject transfer learning) Power of alpha nd beta CSP RCSP Accuracy Method of features extraction Session to Session Subject to Subject 1: Power of alpha and beta 2: CSP 3: RCSP 4:ICA_RCSP Accuracy