Automatic channel selection using multiobjective X-shaped binary butterfly algorithm for motor imagery classification
•A Multi-objective X-shaped Butterfly algorithm is proposed for channel selection.•Our method mimics food foraging behaviour of butterflies in channel selection.•Two Sigmoid transfer functions are used to enhance the solution diversity.•Three public BCI competition datasets are used for validation.•...
Uloženo v:
| Vydáno v: | Expert systems with applications Ročník 206; s. 117757 |
|---|---|
| Hlavní autoři: | , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
Elsevier Ltd
15.11.2022
|
| Témata: | |
| ISSN: | 0957-4174, 1873-6793 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Shrnutí: | •A Multi-objective X-shaped Butterfly algorithm is proposed for channel selection.•Our method mimics food foraging behaviour of butterflies in channel selection.•Two Sigmoid transfer functions are used to enhance the solution diversity.•Three public BCI competition datasets are used for validation.•Our method realized the highest classification accuracy with lesser channels.
Multichannel EEG data processing is usually required to decode Motor Imagery (MI) specific cognitive patterns in Brain-Computer Interface (BCI) systems. The signals from its channels contain information about the underlying neuronal activities that may be redundant and irrelevant to some extent, thereby increasing the computational burden of a BCI system. Moreover, the involvement of additional channels increases the BCI system's hardware complexity, which requires more effort during the BCI preparation setup. Therefore, it is essential to reduce these efforts using a minimal but most informative set of channels. In this study, we developed a Multiobjective X-shaped Binary Butterfly Optimization Algorithm (MX-BBOA) to select the most informative channels from the original set. Firstly, a fifth-order Butterworth bandpass filter is used to collect relevant frequency responses, and then Independent Component Analysis (ICA) is applied to remove artifacts from the filtered signals. The refined signals are further used to extract spatial–temporal features using the Multivariate Empirical Mode Decomposition (MEMD) method. Our approach used an X-shaped transfer function to reduce continuous channel search space to binary search space. The extracted features are used to distinguish multiple MI task pairs such as left hand, right hand, tongue, and feet using the Support Vector Machine (SVM). The experiment is validated on three public EEG datasets (BCI Competition IV- 2008 – IIA, BCI Competition IV- dataset 1, BCI competition III - dataset IVa). The results show that the proposed method achieved a superior classification accuracy (84.49% on dataset 1, 79.74% on dataset 2, and 84.55% on dataset 3) with fewer channels than other state-of-the-art methods. In addition, the computation time compared to other published results was significantly reduced without compromising the classification accuracy. Topographical mapping between the selected channels and the cognitive regions showed that the central, frontal, and parietal lobes execute various MI tasks during physical activities. |
|---|---|
| ISSN: | 0957-4174 1873-6793 |
| DOI: | 10.1016/j.eswa.2022.117757 |