Papers
Topics
Authors
Recent
Search
2000 character limit reached

Emotion Recognition with Machine Learning Using EEG Signals

Published 18 Mar 2019 in cs.LG, cs.HC, and stat.ML | (1903.07272v2)

Abstract: In this research, an emotion recognition system is developed based on valence/arousal model using electroencephalography (EEG) signals. EEG signals are decomposed into the gamma, beta, alpha and theta frequency bands using discrete wavelet transform (DWT), and spectral features are extracted from each frequency band. Principle component analysis (PCA) is applied to the extracted features by preserving the same dimensionality, as a transform, to make the features mutually uncorrelated. Support vector machine (SVM), K-nearest neighbor (KNN) and artificial neural network (ANN) are used to classify emotional states. The cross-validated SVM with radial basis function (RBF) kernel using extracted features of 10 EEG channels, performs with 91.3% accuracy for arousal and 91.1% accuracy for valence, both in the beta frequency band. Our approach shows better performance compared to existing algorithms applied to the "DEAP" dataset.

Citations (86)

Summary

  • The paper demonstrates that SVM with an RBF kernel, combined with wavelet-based feature extraction, achieves classification accuracies of 91.3% for arousal and 91.1% for valence.
  • It employs the DEAP dataset with rigorous preprocessing methods, including AMR normalization and decomposition into five EEG frequency bands.
  • The study highlights the potential of EEG-based emotion recognition to enhance brain-computer interfaces for clinical applications and suggests exploring ensemble strategies.

Emotion Recognition with Machine Learning Using EEG Signals

Introduction

The use of electroencephalography (EEG) signals for emotion recognition has emerged as a significant research area. The paper "Emotion Recognition with Machine Learning Using EEG Signals" (1903.07272) aims to develop an emotion recognition system by leveraging the valence/arousal model through advanced machine learning techniques. The study focuses on classifying emotional states using EEG signals, decomposed into multiple frequency bands, and explores the capabilities of different classifiers within this framework.

Methodology

Data Acquisition and Preprocessing

The research utilizes the DEAP dataset, which is a prominent database for emotion analysis based on physiological signals. The dataset comprises EEG recordings from 32 participants, who were shown 40 one-minute video clips designed to evoke different emotional states. The EEG signals are pre-processed to mitigate noise using the Average Mean Reference (AMR) method and normalized to account for individual variability.

Feature Extraction and Dimensionality Reduction

The core of the feature extraction process involved applying the Discrete Wavelet Transform (DWT) to decompose EEG signals into five frequency bands: theta, alpha, beta, gamma, and noise. Within these bands, the study extracts energy and entropy as spectral features. Principal Component Analysis (PCA) is employed exclusively for decorrelating these features without reducing their dimensionality.

Classification Techniques

For classification, the study evaluates Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel, K-Nearest Neighbor (KNN), and Artificial Neural Network (ANN). The SVM, due to its robustness in creating optimal hyperplanes, particularly excels when dealing with nonlinear data transformations—a characteristic effectively handled by the RBF kernel. Eight-fold cross-validation is implemented to ensure the reliability of the results.

Results

The study reports that among the classifiers, the SVM shows superior performance, achieving a classification accuracy of 91.3% for arousal and 91.1% for valence when trained on features extracted from the beta frequency band. This accuracy significantly surpasses previously reported figures in the literature using the same DEAP dataset, where the maximum reported accuracy was 86.7% using a KNN model.

Discussion and Implications

The findings underscore the effectiveness of utilizing SVM with RBF kernel for emotion classification tasks based on EEG signals, demonstrating improved accuracy over other conventional methods. This approach highlights the importance of feature decorrelation through PCA and meticulous feature extraction procedures exploiting multi-resolution analysis with DWT.

The potential implications of this research are considerable, particularly in enhancing brain-computer interface (BCI) systems for clinical applications, such as monitoring emotional states in patients with psychological disorders. The study suggests further exploration into ensemble learning strategies and the application of other dimension-reducing transformations like Independent Component Analysis (ICA) or Linear Discriminant Analysis (LDA) for possibly elevating classifier performance.

Conclusion

This research presents a substantial advancement in EEG-based emotion recognition, providing a methodological framework that elevates classification accuracy relative to preceding studies. Future research might explore broader frequency decompositions, alternative classification algorithms, and real-time applications to extend the utility of these findings further.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Collections

Sign up for free to add this paper to one or more collections.