Efficient Artifact Identification in Multi-Channel EEG Data

Author/Creator ORCID




Computer Science and Electrical Engineering


Computer Science

Citation of Original Publication


Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu



signals such as Electroencephalogram (EEG) can be used in a variety of purposes including detecting fatigue, stress, brain disorders, brain-computer interfaces (BCIs), or building better models of human variability and human brain. However, EEG signal is frequently contaminated with other sources not related to brain activity. These artifacts may emerge from external sources such as eye blinks, muscle & head movement. In this work, we examine the problem of identifying multiple artifacts on continuous multi-channel EEG data. We first propose a Convolution Neural Networks (CNN) architecture for binary detection of EEG artifact, then further modify the architecture for classifying multiple types of artifacts. The proposed models do not need expert knowledge for feature extraction or pre-processing of EEG data and have a very efficient architecture implementable on mobile devices. We further enhance the architecture to reduce the computation and parameter size through hyperparameter optimization and the use of Depthwise and Separable convolution layers. We propose five different CNN models and evaluate and compare against each other in terms of accuracy, weight parameters, and computation requirements. Our optimized network achieves 94.17% classification accuracy averaged across 17 patients and 9 artifact classes. Compared to the original CNN based architecture, the optimized architecture provides 4.2x and 2.7x less parameters and computation, respectively and has 17.5% higher accuracy. The proposed model was also evaluated on an EEG dataset collected in our lab using a 14-channel Emotiv EPOC headset, and achieves 93.5% accuracy in detecting eye blink artifact.