Fig 1.
A schematic representation of the functioning of onEEGwaveLAD, an online EEG wavelet-based learning adaptive denoising framework.
Fig 2.
High-level diagram of the design of an experiment for evaluating onEEGwaveLAD: A fully automated online EEG wavelet-based learning adaptive denoiser for artefact identification and mitigation.
Table 1.
Parameters of an instantiation of the onEEGwaveLAD pipeline for blink identification and reduction for the Fp1 and Fp2 EEG channels.
Fig 3.
An illustration of coefficient space over eight scales resulting from applying the Discrete Wavelet Transform of a neural signal.
Fig 4.
Distributions of the accuracies for blink identification by the onEEGwaveLAD at each expansion step, grouped by buffer size (column) and the pre-frontal channels (Fp1, Fp2).
Fig 5.
Probability distributions of the means of the signal-to-noise ratios (SNR) for all the EEG windows processed for all the subjects (30), grouped by prediction category (TP = True Positives, TN = True Negatives, FP = False Positives, FN = False Negatives) for a specific instantiation of the onEEGwaveLAD pipeline (EEG Window Length = 1000ms, Sampling rate = 1024, Mother wavelet = Sym4, Buffer capacity = 20, IF sub-sampling size = 512, Number of IF trees = 100, Anomaly Threshold:0.55, Expansion step = 35).