ECIS: EEG Based Classifier for Inner Speech

Introduction: Inner speech, the silent stream of thoughts, offers a novel means of communication and control, particularly for individuals with motor impairments. The proposed EEG-based Communication and Interaction System (ECIS) leverages transfer learning and signal processing techniques to classify inner speech cues, enabling hands-free interactions. This research explores the usability of inner speech in various domains, including healthcare, education, and entertainment, by analyzing its effectiveness in EEG-based command execution. Objectives: This study aims to evaluate the feasibility of inner speech as a reliable communication modality using EEG signals. It seeks to classify directional cues (up, down, left, right) from inner, pronounced, and visualized speech while comparing their recognition accuracy. Additionally, the research examines the scalability of ECIS for more extensive datasets and its potential applications for assistive technologies. Methods: EEG signals were processed through segmentation and transformed using Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) analysis. A transfer learning approach was applied to classify the selected directional cues, and performance was compared across inner, pronounced, and visualized speech conditions. The system was evaluated on a selected dataset, with accuracy comparisons made against existing approaches and simplified cue sets. Results: The model demonstrated the highest classification accuracy for inner speech at 73.05%, outperforming both visualized and pronounced speech. Accuracy comparisons with previous studies using the same cues showed an improvement in recognition rates. Additionally, using a simpler set of cues resulted in variations in accuracy, highlighting the impact of cue complexity on model performance. Conclusions: The findings confirm that inner speech is a viable modality for EEG-based communication, offering significant potential for hands-free interaction. ECIS provides a foundation for future research, particularly in developing scalable models with larger, high-volume datasets. This work paves the way for enhanced human-computer interaction, benefiting individuals with disabilities and expanding applications in various fields.Top of FormBottom of Form