A Novel EEG-Based Deep Learning Framework for Enhancing Communication in Locked-In Syndrome Using P300 Speller and Attention Mechanisms

• Arun Kumar S,
• L Anand,

Vol. 19, No. 11, November 30, 2025
10.3837/tiis.2025.11.006, Download Paper (Free):

• Amyotrophic Lateral Sclerosis
• Convolutional neural networks
• Discrete wavelet transform
• Electroencephalography
• Locked-In Syndrome
• Long short-term memory
• Brain–Computer Interface
• Attention Mechanism

Abstract

This research proposes a deep learning–based communication framework designed for individuals with Locked-In Syndrome (LIS), particularly those in advanced stages of Amyotrophic Lateral Sclerosis (ALS). Using a P300 speller interface, brain signals from 20 participants were recorded as they concentrated on individual characters to construct simple communication phrases. The collected Electroencephalography (EEG) data were processed with Discrete Wavelet Transform (DWT) for feature extraction and then analyzed using an ensemble model that combines Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and an attention mechanism. Experimental results revealed that the proposed model delivered superior performance, achieving 94% Accuracy, 93% Precision, 92% Recall and 92.5% F1-Score clearly outperforming traditional classifiers and baseline deep learning models. These outcomes demonstrate that the framework is both robust and reliable, making it a promising, non-invasive solution to restore communication abilities for patients living with LIS.

Statistics

Cite this article