Eye-Blink Artifact Reduction Using 2-Step Nonnegative Matrix Factorization for Single-Channel Electroencephalographic Signals

Abstract

Artifact reduction from electroencephalographic (EEG) signals is an important process in the numerical analysis of brain activities. In general, independent component analysis (ICA) is employed for artifact reduction from multichannel EEG devices. On the other hand, single-channel EEG devices have recently become attractive because of their usability for measurement and their portability. However, it is ill-defined problem to design a numerical approach for eye-blink artifact reduction from single-channel EEG signals. In this paper, we therefore propose a new artifact reduction method based on 2-step nonnegative matrix factorization (NMF) for single-channel EEG signals. In an experiment, we conducted 2-step NMF to reject eye-blink artifacts using single-channel EEG signals recorded at Fp1. We also applied ICA to multichannel EEG signals and compared the results with those obtained by the proposed method. The experimental results show a relatively high signal-to-noise ratio (SNR) between the signals reconstructed using the proposed method and those obtained by ICA. Moreover, we confirm the coefficient of correlation of over 99% for estimating the recorded EEG signals by the proposed method.

An automatic algorithm for blink-artifact suppression based on iterative template matching: application to single channel recording of cortical auditory evoked potentials

“…This paper also questioned the blink-artifact triangular shape assumption considered by some techniques like MSDW. The blink-artifact waveforms observed across subjects in this study presented a similar morphology, consistent with previous studies [67,68]: an onset with an abrupt peak following a lower frequency component which extends up to 1 second approximately. Figure 5.A shows an EEG segment contaminated by a number of blink-artifacts (RAW) and the same segment corrected by the MSDW and ITMS methods.…”

“…This approach was followed because, despite other techniques having previously attempted to suppress blink-artifacts in single EEG channel applications, we did not find any alternative satisfactory option. Kanoga and Mitsukura (2014) described a method suppressing from the EEG the frequency components of the blink-events, which are estimated through a two-step non-negative matrix factorization [67]. However, this method does not allow blink-events detection, and its performance is strongly dependent on the selection of the basis K 1 and K 2 , which requires human intervention and with no procedure available to obtain an optimal selection.…”

Independent Low-Rank Matrix Analysis-Based Automatic Artifact Reduction Technique Applied to Three BCI Paradigms

“…, a iQ ] H ∈ C P×Q is a frequency-wise mixing matrix (a iq is the steering vector for the qth source, and H indicates the Hermitian transpose). In this paper, we set the value of time length to 1 s because some frequency-domain artifact reduction techniques translate EEG data into STFT domain based on 1-s windows (Kanoga and Mitsukura, 2014;Mohammadpour and Rahmani, 2017) and ILRMA showed its high separation accuracy for 1-s time length data (Kitamura et al, 2017). This mixing system is the rank-1 spatial model (Duong et al, 2010); thus, the relationships between observations and sources can be represented:ŝ…”

“…Recent studies attempt to propose a generic artifact removal algorithm (Chen et al, 2019). Unlike the time-domain algorithm, frequency-domain methods (i.e., IVA and ILRMA) can separate single-channel data if the differences in data-driven spectral basis functions can be learned well (Kanoga and Mitsukura, 2014;Kanoga et al, 2019a). Thus, we will investigate our proposed algorithm in practical situations and extend it as a generic and user-friendly algorithm for reducing artifacts from EEG data.…”

Review of Artifact Rejection Methods for Electroencephalographic Systems

“…Some works reported that the supervised NMF could effectively factorize the observed EEG signals into the brain activity components and the artifacts if the user has artifact data in advance [62,63]. Before applying supervised learning, template matrix X Art has been factorized into H Art and W Art .ThematrixX is continuously factorized into H and W where H contains the elements of matrix H Art .…”