The Role of Feature Selection in Significant Information Extraction from EEG Signals

Abstract

Information extraction from EEG signals for use in Brain Machine Interface systems has been a highly effective research topic recently. Due to the complexity, high dimension, and subject specific behavior of the EEG signals make feature extraction and selection very important. For this reason, there are many studies in the direction of feature extraction and selection which affect the performance of the Brain Machine Interface system at a high level. In this study, different statistical characteristics were obtained from wavelet coefficients obtained by wavelet transform by using BCI Competition IV-2b data set. The selection of the efficient ones of these features is provided by Principal Component Analysis. The fitness of logistic regression model established with both feature groups was measured by Akaike Information Criteria. The results indicated that relatively better statistical performance can be obtained by using fewer features thanks to PCA. These results are important in terms of statistical comparison and demonstration of the success in extracting information from EEG signals.

Keywords

• EEG
• Brain Machine Interface
• Feature Extraction
• Feature Selection
• Akaike Information Criteria

References

1. P. K. Pattnaik and J. Sarraf, “Brain Computer Interface issues on hand movement,” J. King Saud Univ. Inf. Sci., vol. 30, no. 1, pp. 18–24, 2018.
2. M. Fatourechi, A. Bashashati, R. K. Ward, and G. E. Birch, “EMG and EOG artifacts in brain computer interface systems: A survey,” Clin. Neurophysiol., vol. 118, no. 3, pp. 480–494, 2007.
3. N. S. Malan and S. Sharma, “Feature selection using regularized neighbourhood component analysis to enhance the classification performance of motor imagery signals,” Comput. Biol. Med., vol. 107, pp. 118–126, 2019.
4. S. Aggarwal and N. Chugh, “Signal processing techniques for motor imagery brain computer interface: A review,” Array, vol. 1, p. 100003, 2019.
5. R. Leeb, C. Brunner, G. Müller-Putz, A. Schlögl, and G. Pfurtscheller, “BCI Competition 2008–Graz data set B,” Graz Univ. Technol. Austria, pp. 1–6, 2008.
6. K. Burnham and D. Anderson, “Model Selection and Multimodel Inference,” Technometrics, vol. 45, pp. 181–181, 2003.
7. M. Yang, Y.-F. Sang, C. Liu, and Z. Wang, “Discussion on the choice of decomposition level for wavelet based hydrological time series modeling,” Water, vol. 8, no. 5, p. 197, 2016.
8. Raza, H., H. Cecotti, Y. Li and G. Prasad, “Adaptive learning with covariate shift-detection for motor imagery-based brain–computer interface,”. Soft Computing, vol. 20,no. 8, pp. 3085–3096, 2016.

9. Sayed, K., M. Kamel, M. Alhaddad, H.M. Malibary and Y.M. Kadah, “Characterization of phase space trajectories for Brain-Computer Interface,” Biomedical Signal Processing and Control, vol.38, pp.55–66, 2017.
10. R. K. Chaurasiya, N. D. Londhe, and S. Ghosh, “Statistical wavelet features, PCA, and SVM based approach for EEG signals classification,” Int. J. Electr. Comput. Electron. Commun. Eng., vol. 9, no. 2, pp. 182–186, 2015.
11. E. Dagdevir and M. Tokmakci, “Determination of Effective Signal Processing Stages for Brain Computer Interface on BCI Competition IV Data Set 2b: A Review Study,” IETE Journal of Research,1914204, 2021.
12. E. Dagdevir, M. Kocaturk, and M. Okatan, “Likelihood-Based Amplitude Thresholding in Extracellular Neural Recordings,” 27th Signal Processing and Communications Applications Conference, 2019, pp. 1–4.