Premium
A background EEG removal method combining PCA with multivariate empirical mode decomposition for event‐related potential measurements
Author(s) -
Kawaguchi Hirokazu,
Kume Takahiro,
Kobayashi Tetsuo
Publication year - 2013
Publication title -
ieej transactions on electrical and electronic engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.254
H-Index - 30
eISSN - 1931-4981
pISSN - 1931-4973
DOI - 10.1002/tee.21918
Subject(s) - principal component analysis , electroencephalography , multivariate statistics , hilbert–huang transform , oddball paradigm , pattern recognition (psychology) , computer science , event related potential , artificial intelligence , noise (video) , speech recognition , independent component analysis , white noise , machine learning , psychology , image (mathematics) , telecommunications , psychiatry
The event‐related potential (ERP) is a neural response to an internal or external event, and can be obtained by averaging time‐locked scalp potentials. The ERP measured in a single trial often has a low signal‐to‐noise ratio (SNR) because of the relatively large background due to the rhythmic electroencephalogram (EEG) noise. This paper proposes a novel method to enhance ERPs by combining principal component analysis (PCA) with multivariate empirical mode decomposition (M‐EMD). EMD is a data‐driven time–frequency analysis of nonlinear and nonstationary signals, and M‐EMD is its multivariate extension. In the proposed method, PCA reduces the data dimensions, while M‐EMD removes the relatively large background EEGs. The performance of the method is evaluated with simulated and measured P300 ERP components obtained from a visual oddball experiment. The results demonstrate that the proposed method can substantially reduce the background EEGs and improve the SNR of P300s. © 2013 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.