TY - GEN
T1 - Removal of eye-blinking artifacts by ICA in cross-modal long-term EEG recording
AU - Huang, Gan
AU - Hu, Zhenxing
AU - Zhang, Li
AU - Li, Linling
AU - Liang, Zhen
AU - Zhang, Zhiguo
N1 - Publisher Copyright:
© 2020 IEEE.
PY - 2020/7
Y1 - 2020/7
N2 - Independent Component Analysis (ICA) has became the most popular method to remove eye-blinking artifacts from electroencephalogram (EEG) recording. For long term EEG recording, ICA was commonly considered to costing a lot of computation time. Furthermore, with no ground truth, the discussion about the quality of ICA decomposition in a nonstationary environment was specious. In this study, we investigated the signal (P300 waveform) and the noise (averaged eye-blinking artifacts) on a cross-modal long-term EEG recording to evaluate the efficiency and effectiveness of different methods on ICA eye-blinking artifacts removal. As a result, it was found that, firstly, down sampling is an effective way to reduce the computation time in ICA. Appropriate down sampling ratio could speed up ICA computation 200 times and keep the decomposition performance stable, in which the computation time of ICA decomposition on a 2800 s EEG recording was less than 5 s. Secondly, dimension reduction by PCA was also a way to improve the efficiency and effectiveness of ICA. Finally, the comparison by cropping the dataset indicated that performing ICA on each run of the experiment separately would achieve a better result for eye-blinking artifacts removal than using all the EEG data input for ICA.
AB - Independent Component Analysis (ICA) has became the most popular method to remove eye-blinking artifacts from electroencephalogram (EEG) recording. For long term EEG recording, ICA was commonly considered to costing a lot of computation time. Furthermore, with no ground truth, the discussion about the quality of ICA decomposition in a nonstationary environment was specious. In this study, we investigated the signal (P300 waveform) and the noise (averaged eye-blinking artifacts) on a cross-modal long-term EEG recording to evaluate the efficiency and effectiveness of different methods on ICA eye-blinking artifacts removal. As a result, it was found that, firstly, down sampling is an effective way to reduce the computation time in ICA. Appropriate down sampling ratio could speed up ICA computation 200 times and keep the decomposition performance stable, in which the computation time of ICA decomposition on a 2800 s EEG recording was less than 5 s. Secondly, dimension reduction by PCA was also a way to improve the efficiency and effectiveness of ICA. Finally, the comparison by cropping the dataset indicated that performing ICA on each run of the experiment separately would achieve a better result for eye-blinking artifacts removal than using all the EEG data input for ICA.
UR - https://www.scopus.com/pages/publications/85091032293
U2 - 10.1109/EMBC44109.2020.9176711
DO - 10.1109/EMBC44109.2020.9176711
M3 - 会议稿件
AN - SCOPUS:85091032293
T3 - Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS
SP - 217
EP - 220
BT - 42nd Annual International Conferences of the IEEE Engineering in Medicine and Biology Society
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 42nd Annual International Conferences of the IEEE Engineering in Medicine and Biology Society, EMBC 2020
Y2 - 20 July 2020 through 24 July 2020
ER -