Optimized deformable convolution network for detection and mitigation of ocular artifacts from EEG signal
Devulapalli Shyam Prasad1, Srinivasa Rao Chanamallu2, Kodati Satya Prasad3
1JNTU, Kakinada / CVR College of Engineering, Hyderabad, Telangana India.
Multimedia Tools and Applications
|April 18, 2022
Summary
This study introduces a new deep learning model to detect and remove ocular artifacts from electroencephalogram (EEG) signals. The advanced method effectively cleans EEG data, improving brain activity analysis for applications like Brain-Computer Interface.
Area of Science:
- Neuroscience and Biomedical Engineering
- Signal Processing and Machine Learning
Background:
- Electroencephalogram (EEG) signals are crucial for analyzing brain activity but are often contaminated by artifacts.
- Ocular artifacts, originating from eye movements, are a significant source of interference in EEG recordings.
- Effective artifact removal is essential for accurate EEG analysis, especially in real-time applications like Brain-Computer Interfaces (BCI).
Purpose of the Study:
- To develop and validate a novel deep learning-based model for the online detection and prevention of ocular artifacts in EEG signals.
- To enhance the efficiency and accuracy of artifact removal processes in EEG recordings.
- To improve the quality of EEG data for reliable brain activity and behavior analysis.
Main Methods:
- Signal decomposition using 5-level Discrete Wavelet Transform (DWT) and Pisarenko harmonic decomposition.
- Feature extraction via Principle Component Analysis (PCA) and Independent Component Analysis (ICA).
- Ocular artifact detection using optimized Deformable Convolutional Networks (DCN) tuned by Distance Sorted-Electric Fish Optimization (DS-EFO), followed by Empirical Mean Curve Decomposition (EMCD) for artifact mitigation and signal denoising.
Main Results:
- The proposed deep learning model demonstrated superior performance in ocular artifact detection and removal compared to conventional methods.
- Achieved higher accuracy in identifying and mitigating ocular artifacts from diverse EEG datasets.
- Successfully generated clean EEG signals through an online artifact prevention and denoising process.
Conclusions:
- The developed deep learning-based model offers an effective solution for ocular artifact reduction in EEG signals.
- The proposed method significantly improves the quality of EEG data, benefiting applications requiring precise brain signal analysis.
- This approach holds promise for real-time artifact management in BCI and other neurophysiological monitoring systems.
Keywords:
5-level discrete wavelet transformDistance sorted-electric fish optimizationElectroencephalogramEmpirical mean curve decompositionOcular artifactsOptimized deformable convolutional networksPisarenko harmonic decomposition
