Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

How Homogenizing the Channel-wise Magnitude Can Enhance EEG Classification Model?

Huyen Ngo, Khoi Do, Duong Nguyen, Viet Dung Nguyen, Lan Dang

TL;DR

The work tackles EEG classification challenges arising from multi-channel redundancy by introducing Inverted Channel-wise Magnitude Homogenization (ICWMH) to equalize channel contributions and a Feature Enrichment via Skip Connection (FEvSC) that leverages edge detection to enrich representations. EEG signals are converted into encoded images and processed through a CNN with skip connections, enabling effective feature fusion via Hadamard-like fusion. Empirical results on the Perceive Lab 40-class and High-Gamma datasets show the approach achieving approximately 66% and 57.2% accuracy, respectively, outperforming several baselines and demonstrating the value of targeted preprocessing for improving EEG classification with lighter networks. The method highlights the importance of balanced channel input and edge-focused feature enrichment, with ablation studies guiding hyperparameter choices for interpolation, thresholds, and smoothing.

Abstract

A significant challenge in the electroencephalogram EEG lies in the fact that current data representations involve multiple electrode signals, resulting in data redundancy and dominant lead information. However extensive research conducted on EEG classification focuses on designing model architectures without tackling the underlying issues. Otherwise, there has been a notable gap in addressing data preprocessing for EEG, leading to considerable computational overhead in Deep Learning (DL) processes. In light of these issues, we propose a simple yet effective approach for EEG data pre-processing. Our method first transforms the EEG data into an encoded image by an Inverted Channel-wise Magnitude Homogenization (ICWMH) to mitigate inter-channel biases. Next, we apply the edge detection technique on the EEG-encoded image combined with skip connection to emphasize the most significant transitions in the data while preserving structural and invariant information. By doing so, we can improve the EEG learning process efficiently without using a huge DL network. Our experimental evaluations reveal that we can significantly improve (i.e., from 2% to 5%) over current baselines.

How Homogenizing the Channel-wise Magnitude Can Enhance EEG Classification Model?

TL;DR

The work tackles EEG classification challenges arising from multi-channel redundancy by introducing Inverted Channel-wise Magnitude Homogenization (ICWMH) to equalize channel contributions and a Feature Enrichment via Skip Connection (FEvSC) that leverages edge detection to enrich representations. EEG signals are converted into encoded images and processed through a CNN with skip connections, enabling effective feature fusion via Hadamard-like fusion. Empirical results on the Perceive Lab 40-class and High-Gamma datasets show the approach achieving approximately 66% and 57.2% accuracy, respectively, outperforming several baselines and demonstrating the value of targeted preprocessing for improving EEG classification with lighter networks. The method highlights the importance of balanced channel input and edge-focused feature enrichment, with ablation studies guiding hyperparameter choices for interpolation, thresholds, and smoothing.

Abstract

A significant challenge in the electroencephalogram EEG lies in the fact that current data representations involve multiple electrode signals, resulting in data redundancy and dominant lead information. However extensive research conducted on EEG classification focuses on designing model architectures without tackling the underlying issues. Otherwise, there has been a notable gap in addressing data preprocessing for EEG, leading to considerable computational overhead in Deep Learning (DL) processes. In light of these issues, we propose a simple yet effective approach for EEG data pre-processing. Our method first transforms the EEG data into an encoded image by an Inverted Channel-wise Magnitude Homogenization (ICWMH) to mitigate inter-channel biases. Next, we apply the edge detection technique on the EEG-encoded image combined with skip connection to emphasize the most significant transitions in the data while preserving structural and invariant information. By doing so, we can improve the EEG learning process efficiently without using a huge DL network. Our experimental evaluations reveal that we can significantly improve (i.e., from 2% to 5%) over current baselines.
Paper Structure (14 sections, 8 equations, 2 figures, 2 tables)

This paper contains 14 sections, 8 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Preliminary and Notations
  4. Overall Architecture
  5. Inverted Channel-wise Magnitude Homogenization (ICWMH)
  6. Variant Feature Extractor
  7. Feature Enrichment via Skip Connection (FEvSC)
  8. Experiment
  9. Comparison to State-of-the-art methods
  10. Ablation Study
  11. Conclusion
  12. Related Works
  13. Model Architecture Design
  14. Feature Enhancement

Figures (2)

  • Figure 1: Overall two-step preprocessing methodology for EEG signal classification. 1) ICWMH: an input EEG signal $x^i$ fed into the ICWMH process to normalize channel amplitudes and generate an encoded image with the size of $C \times L$ of the sample. 2) FEvSC: FEvSC uses edge detection to extract useful variation information from an image. This extracted data is subsequently incorporated back into the encoded image. This potentially improves the performance of the EEG classification task.
  • Figure 2: Overview of Variant Feature Extractor Method