Content Details | NUML Online Research Repository

Title EEG-BASED EMOTION ANALYSIS USING TEXT STIMULI

Author(s) IHSAN ULLAH

Abstract Emotions are fundamental to our daily lives, impacting our behavior, thoughts, and feelings. Accurately classifying emotions is of utmost importance in various fields, including psychology, psychiatry, and neuroscience, as it can aid in the development of effective diagnostic tools, and human-computer interaction systems. In this study, the aim was to classify emotions using Electroencephalogram (EEG) data, which provides a non-invasive and objective measure of brain activity. Data was collected from 25 participants using a 128-channel device and text stimuli. To achieve this goal, a comprehensive approach was adopted that integrates multiple feature extraction techniques, pre-processing techniques, and a Support Vector Machine (SVM) classifier. Four feature extraction techniques, Convolutional Neural Network (CNN), Wavelet Transform (WT), Power Spectral Density (PSD), and Raw data, are used to extract features from pre-processed EEG signals. The pre-processing techniques involved down sampling, re-referencing, and filtering the EEG signals to eliminate noise and artifacts. The Monte Carlo approach is applied to randomly selecting training and testing samples to ensure the reliability and validity of results of this study. Study focused on classifying emotions as positive and negative. T-tests were used to identify the most relevant features that contributed to the classification of emotions. The results show that the combination of CNN and SVM yields the highest average accuracy rate of 80%, followed by WT with 75%, PSD with 72%, and raw data with 65%. This suggests that the use of CNN, WT, and PSD as feature extraction techniques in combination with SVM as a classifier can significantly improve the classification of emotions based on EEG data. The proposed approach has significant implications for the development of more accurate and efficient methods for classifying emotions in various fields. For instance, in the field of human-computer interaction, accurate emotion recognition can be used to develop personalized interfaces that respond to the user's emotional state. In clinical settings, the accurate classification of emotions can aid in the diagnosis of psychiatric disorders and inform treatment strategies. Our findings highlight the potential of EEG data as a valuable source of information for emotion classification, with important applications in various domains.

Type Thesis/Dissertation MS

Faculty Engineering and Computer Science

Department Engineering

Language English

Publication Date 2023-06-02

Subject

Publisher

Contributor(s)

Format

Identifier

Source

Relation

Coverage

Rights