作者: Azin S. Janani , Tyler S. Grummett , Hanieh Bakhshayesh , Trent W. Lewis , John O. Willoughby
DOI: 10.23919/EUSIPCO.2018.8553261
关键词: Artificial intelligence 、 Pattern recognition 、 Independent component analysis 、 Scalp 、 Eeg recording 、 Electromyography 、 Tonic (physiology) 、 Computer science 、 Neurophysiology 、 Electroencephalography
摘要: Scalp electrical recordings, or electroencephalograms (EEG), are heavily contaminated by cranial and cervical muscle activity from as low 20 hertz, even in relaxed conditions. It is therefore necessary to reduce remove this contamination enable reliable exploration of brain neurophysiological responses. measurements record many sources, including neural muscular. Independent Component Analysis (ICA) produces components ideally corresponding separate but the number limited EEG channels. In practice, at most 30% cleanly sources. Increasing channels results more components, with a significant increase costs data collection computation. Here we present assist selecting an appropriate Our unique database pharmacologically paralysed subjects provides way objectively compare different approaches achieving ideal, free recording. We evaluated automatic muscle-removing approach, based on ICA, numbers channels: 21, 32, 64, 115. show that, for fixed length data, 21 insufficient tonic artefact, that increasing 115 does result better artefact reduction.