EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Classification of Envisioned English Speech from EEG using deep learning approaches

Arman Hossain, Tanvir Hossain Ovi, Mohammed Arif Iftakher Mahmood, Md. Fazlul Kader

Machine Learning with Applications · 2025-12 · 22 · 100752

DOI: 10.1016/j.mlwa.2025.100752 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Machine Learning with Applications, 2025-12. Verfasst von Arman Hossain und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Information and Communication Technology Division.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The ability to use non-invasive technology to recognize imagined speech holds significant potential to enhance the well-being of individuals with motor neuron disease, especially when voluntary muscle control is limited. Despite its potential, this technology is still in its early stages of development, and there remains a significant scope for further research. Therefore, we proposed deep learning approaches to classify 26 English alphabets and 10 digits. To train and evaluate the models, we have collected electroencephalogram (EEG) signals from 20 participants in real-time while they imagined the alphabet and digits. The raw EEG data were processed using the filters, which allowed us to separate various frequency bands: δ , θ , α , β , and γ , as well as combinations of these bands δ + β + γ , and θ + α + β . The dataset was divided into training (80%), validation (10%), and testing (10%) sets and evaluated across three deep learning models: 1D CNN, BiLSTM, and a hybrid 1D CNN–BiLSTM. Among them, the BiLSTM model outperformed the other two models, achieving the highest test accuracy of 85.65% in digit classification within the δ frequency band and 83.65% in alphabet classification within the α frequency band. This research has the potential to facilitate the development of assistive communication technologies for individuals with motor neuron disorders.

Förderung

  • Information and Communication Technology Division

Zitieren

APA

Hossain, A., Ovi, T. H., Mahmood, M. A. I., & Kader, M. F. (2025). Classification of Envisioned English Speech from EEG using deep learning approaches. Machine Learning with Applications, 22, 100752. https://doi.org/10.1016/j.mlwa.2025.100752

Vancouver

Hossain A, Ovi TH, Mahmood MAI, Kader MF. Classification of Envisioned English Speech from EEG using deep learning approaches. Machine Learning with Applications. 2025;22:100752. doi:10.1016/j.mlwa.2025.100752

BibTeX
@article{Hossain2025,
  author = {Hossain, Arman and Ovi, Tanvir Hossain and Mahmood, Mohammed Arif Iftakher and Kader, Md. Fazlul},
  title = {Classification of Envisioned English Speech from EEG using deep learning approaches},
  journal = {Machine Learning with Applications},
  year = {2025},
  volume = {22},
  pages = {100752},
  doi = {10.1016/j.mlwa.2025.100752}
}
Kostenlos starten →