Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Emotion Detection from EEG using Transfer Learning

Sidharth Sidharth, Ashish Abraham Samuel, Ranjana H, Jerrin Thomas Panachakel, Sana Parveen K

TL;DR

The paper tackles EEG-based emotion detection under data scarcity by applying transfer learning with a ResNet-50 model to a novel image representation of coherence features. Specifically, it fuses Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) across alpha, beta, and gamma bands, with Differential Entropy (DE) added to diagonals, encoded as 62×62 RGB images for classification. Evaluations on SEED_EEG show 93.1% subject-dependent accuracy and 71.6% LOSO subject-independent accuracy, both well above chance for three classes, outperforming several prior methods. The work contributes a principled feature fusion and image-based transfer-learning pipeline with potential implications for EEG-based neuroprosthetics and emotion-aware systems.

Abstract

The detection of emotions using an Electroencephalogram (EEG) is a crucial area in brain-computer interfaces and has valuable applications in fields such as rehabilitation and medicine. In this study, we employed transfer learning to overcome the challenge of limited data availability in EEG-based emotion detection. The base model used in this study was Resnet50. Additionally, we employed a novel feature combination in EEG-based emotion detection. The input to the model was in the form of an image matrix, which comprised Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) in the upper-triangular and lower-triangular matrices, respectively. We further improved the technique by incorporating features obtained from the Differential Entropy (DE) into the diagonal, which previously held little to no useful information for classifying emotions. The dataset used in this study, SEED EEG (62 channel EEG), comprises three classes (Positive, Neutral, and Negative). We calculated both subject-independent and subject-dependent accuracy. The subject-dependent accuracy was obtained using a 10-fold cross-validation method and was 93.1%, while the subject-independent classification was performed by employing the leave-one-subject-out (LOSO) strategy. The accuracy obtained in subject-independent classification was 71.6%. Both of these accuracies are at least twice better than the chance accuracy of classifying 3 classes. The study found the use of MSC and MPC in EEG-based emotion detection promising for emotion classification. The future scope of this work includes the use of data augmentation techniques, enhanced classifiers, and better features for emotion classification.

Emotion Detection from EEG using Transfer Learning

TL;DR

The paper tackles EEG-based emotion detection under data scarcity by applying transfer learning with a ResNet-50 model to a novel image representation of coherence features. Specifically, it fuses Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) across alpha, beta, and gamma bands, with Differential Entropy (DE) added to diagonals, encoded as 62×62 RGB images for classification. Evaluations on SEED_EEG show 93.1% subject-dependent accuracy and 71.6% LOSO subject-independent accuracy, both well above chance for three classes, outperforming several prior methods. The work contributes a principled feature fusion and image-based transfer-learning pipeline with potential implications for EEG-based neuroprosthetics and emotion-aware systems.

Abstract

The detection of emotions using an Electroencephalogram (EEG) is a crucial area in brain-computer interfaces and has valuable applications in fields such as rehabilitation and medicine. In this study, we employed transfer learning to overcome the challenge of limited data availability in EEG-based emotion detection. The base model used in this study was Resnet50. Additionally, we employed a novel feature combination in EEG-based emotion detection. The input to the model was in the form of an image matrix, which comprised Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) in the upper-triangular and lower-triangular matrices, respectively. We further improved the technique by incorporating features obtained from the Differential Entropy (DE) into the diagonal, which previously held little to no useful information for classifying emotions. The dataset used in this study, SEED EEG (62 channel EEG), comprises three classes (Positive, Neutral, and Negative). We calculated both subject-independent and subject-dependent accuracy. The subject-dependent accuracy was obtained using a 10-fold cross-validation method and was 93.1%, while the subject-independent classification was performed by employing the leave-one-subject-out (LOSO) strategy. The accuracy obtained in subject-independent classification was 71.6%. Both of these accuracies are at least twice better than the chance accuracy of classifying 3 classes. The study found the use of MSC and MPC in EEG-based emotion detection promising for emotion classification. The future scope of this work includes the use of data augmentation techniques, enhanced classifiers, and better features for emotion classification.
Paper Structure (8 sections, 4 equations, 4 figures, 2 tables)

This paper contains 8 sections, 4 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Dataset
  3. Feature extraction and Methodology
  4. Experiments
  5. Results
  6. conclusion
  7. Future scope
  8. Ethics Statement

Figures (4)

  • Figure 1: Method used for generating the dataset which is fed to the Classifiers.(BPF refers to band pass filter)
  • Figure 2: Transfer learning approach using ResNet-50 for subject-independent classification
  • Figure 3: Transfer learning approach using ResNet-50 for subject-dependent classification
  • Figure 4: Accuracy for subject-dependent and subject-independent classifications