EEG-MI Training Elimination

Disciplines

Bioelectrical and Neuroengineering | Biomedical | Signal Processing | Systems and Integrative Engineering

Abstract (300 words maximum)

There has been a surge of growing interest in the Artificial Intelligence (AI) field, particularly Machine Learning (ML) techniques, across a wide range of domains lately. Biomedical engineering is one of the disciplines trying to leverage ML capabilities to analyze time-series biological signals for use in the development of assistive technology applications. Of those, Electroencephalography (EEG) signals are brain signals extracted from the brain's cortical location and indicate some of the reactions that occur within the brain as a result of performing mental tasks. Motor Imagery (MI) is a well-known mental activity that is triggered by imagining oneself performing a certain physical task. A major drawback of using the MI paradigm in engineering applications is its dependence on the potential user; in order for the ML model to function properly, the same user who will use the final application will need to perform training sessions before the model is constructed. These sessions resemble a burden on the user and prevent the technology from being used in a practical manner as it cannot be used in a plug-and-play fashion. Thus, this project investigates the possibility of replacing the training sessions with short segments of EEG signals collected at rest after the model is built. This has the potential to transfer several applications from the research space to solve real-world problems.

Academic department under which the project should be listed

Electrical and Computer Engineering

Primary Investigator (PI) Name

Sylvia Bhattacharya

This document is currently not available here.

Share

COinS
 

EEG-MI Training Elimination

There has been a surge of growing interest in the Artificial Intelligence (AI) field, particularly Machine Learning (ML) techniques, across a wide range of domains lately. Biomedical engineering is one of the disciplines trying to leverage ML capabilities to analyze time-series biological signals for use in the development of assistive technology applications. Of those, Electroencephalography (EEG) signals are brain signals extracted from the brain's cortical location and indicate some of the reactions that occur within the brain as a result of performing mental tasks. Motor Imagery (MI) is a well-known mental activity that is triggered by imagining oneself performing a certain physical task. A major drawback of using the MI paradigm in engineering applications is its dependence on the potential user; in order for the ML model to function properly, the same user who will use the final application will need to perform training sessions before the model is constructed. These sessions resemble a burden on the user and prevent the technology from being used in a practical manner as it cannot be used in a plug-and-play fashion. Thus, this project investigates the possibility of replacing the training sessions with short segments of EEG signals collected at rest after the model is built. This has the potential to transfer several applications from the research space to solve real-world problems.