Thought To Text: An EEG Driven Approach For Predicting Letters
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Department
CCSE - Computer Science
Abstract
Brain-Computer Interfaces (BCIs) provide an innovative means of translating the brain's activity into practical data. When the brain activates clusters of neurons to perform a task, these groups of neurons fire together, generating detectable electrical signals. These signals are captured through electroencephalogram (EEG) recordings, representing the brain's electrical activity, or brainwaves. These EEG recordings help capture human thought, enabling communication for individuals who may have lost the ability to communicate through traditional means. Current BCIs face challenges such as invasive data collection, dependency on individual users, and limited text generation methods. Moreover, these approaches may not be feasible for large-scale deployment due to their invasive nature, multi-user signal variation, signal variability of the same user, and noisy external environment. Our research proposes a novel approach, capturing minimal EEG signals through non-invasive methods in conjunction with machine learning algorithms such as Support Vector Machines to classify and predict letters from the English alphabet based on EEG signals. This methodology aims to overcome existing limitations, paving the way for scalable and user-friendly BCIs.
Disciplines
Artificial Intelligence and Robotics | Computer Sciences
Thought To Text: An EEG Driven Approach For Predicting Letters
Brain-Computer Interfaces (BCIs) provide an innovative means of translating the brain's activity into practical data. When the brain activates clusters of neurons to perform a task, these groups of neurons fire together, generating detectable electrical signals. These signals are captured through electroencephalogram (EEG) recordings, representing the brain's electrical activity, or brainwaves. These EEG recordings help capture human thought, enabling communication for individuals who may have lost the ability to communicate through traditional means. Current BCIs face challenges such as invasive data collection, dependency on individual users, and limited text generation methods. Moreover, these approaches may not be feasible for large-scale deployment due to their invasive nature, multi-user signal variation, signal variability of the same user, and noisy external environment. Our research proposes a novel approach, capturing minimal EEG signals through non-invasive methods in conjunction with machine learning algorithms such as Support Vector Machines to classify and predict letters from the English alphabet based on EEG signals. This methodology aims to overcome existing limitations, paving the way for scalable and user-friendly BCIs.