Real-Time EEG Signal Artifact Removal using Independent Component Analysis and its variants
Barath Parthiban, Vishwa Priya, G. Kavitha · 2024
Electroencephalogram (EEG) is the recording of electrical activity of the brain. The acquired EEG signal is corrupted by other physiological noises particularly ocular artifacts. EEG signal filtering is necessary in order to obtain a clean waveform that has potential application in the field of clinical diagnosis and research. Independent component analysis (ICA) is the most commonly used method to remove ocular artifacts. The proposed work uses various types of ICA algorithms that include Fast ICA, Infomax ICA, Extended Infomax ICA, and Picard ICA. Publicly available EEG signal datasets and real-time acquired EEG signals are used. Evaluation metrics PSNR, SNR, and spectral distortion are used to evaluate the performance of different types of ICA algorithms. The results indicate that for Infomax ICA, the PSNR values are 27.32 dB for the public dataset and 24.3 dB for the real-time signal. The SNR values are 16.27 dB for the public dataset and 15.26 dB for the real-time signal. In terms of spectral distortion, the public dataset shows a value of 1.7 dB, and the real-time signal has a value of 1.8 dB. Infomax ICA outperforms other methods in EOG artifact removal with minimal loss of data. This artifact-removed EEG can then be used in applications like brain-computer interface systems, neurological disorder diagnosis, and cognitive neuroscience research.