Automated Thought-to-Text Conversion Through Automated Brainwave Signal Annotation
Disciplines
Computer Sciences
Abstract (300 words maximum)
Brain-Computer Interface (BCI) is a technology that bridges the gap in communication between the brain and external devices. When the brain activates clusters of neurons to perform a task, these groups of neurons fire together and generate a detectable electrical signal. These signals are captured through electroencephalogram (EEG), and the brain's electric activity (brainwave) signals help capture human thoughts to enable communication with people who have lost communication ability. Recent research focuses on predicting letters from invasive-brainwave signals (EEG) by transplanting EEG electrodes in the human brain. These predicted letters could be used to develop thought-to-text applications later. However, building a highly accurate thought-to-text application requires automatically segmenting continuous EEG signals and labeling each segment with a correct English alphabet letter. This work investigates the real-time segmenting and annotating of the brainwave signals with the corresponding English alphabet using non-invasive brainwave signals collected using a 14-channel Emotive headset. Emotive headset-collected brainwave signals are noisy due to the external surface-level contact with the headset; therefore, we filter these signals to remove noise and extract meaningful features that capture insights from the signal. We deploy a lightweight machine learning algorithm (e.g., support vector machine) to identify English alphabet letters in real time and display them in a user interface (UI) for a particular EEG signal segment. Our current approach detects a subset of the English alphabet with 64% accuracy. We plan to improve this current algorithm and develop a novel algorithm to detect words from the detected sequence of letters in the future. Our project assists individuals in communicating with others in the form of writing who suffered any brain damage.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Automated Thought-to-Text Conversion Through Automated Brainwave Signal Annotation
Brain-Computer Interface (BCI) is a technology that bridges the gap in communication between the brain and external devices. When the brain activates clusters of neurons to perform a task, these groups of neurons fire together and generate a detectable electrical signal. These signals are captured through electroencephalogram (EEG), and the brain's electric activity (brainwave) signals help capture human thoughts to enable communication with people who have lost communication ability. Recent research focuses on predicting letters from invasive-brainwave signals (EEG) by transplanting EEG electrodes in the human brain. These predicted letters could be used to develop thought-to-text applications later. However, building a highly accurate thought-to-text application requires automatically segmenting continuous EEG signals and labeling each segment with a correct English alphabet letter. This work investigates the real-time segmenting and annotating of the brainwave signals with the corresponding English alphabet using non-invasive brainwave signals collected using a 14-channel Emotive headset. Emotive headset-collected brainwave signals are noisy due to the external surface-level contact with the headset; therefore, we filter these signals to remove noise and extract meaningful features that capture insights from the signal. We deploy a lightweight machine learning algorithm (e.g., support vector machine) to identify English alphabet letters in real time and display them in a user interface (UI) for a particular EEG signal segment. Our current approach detects a subset of the English alphabet with 64% accuracy. We plan to improve this current algorithm and develop a novel algorithm to detect words from the detected sequence of letters in the future. Our project assists individuals in communicating with others in the form of writing who suffered any brain damage.