Available at: https://digitalcommons.calpoly.edu/theses/3052
Date of Award
6-2025
Degree Name
MS in Electrical Engineering
Department/Program
Electrical Engineering
College
College of Engineering
Advisor
Xiao-Hua Yu
Advisor Department
Electrical Engineering
Advisor College
College of Engineering
Abstract
Emotion recognition using ECG signals has gained traction due to its potential in applications like healthcare, human-computer interaction, and the growing availability of wearable ECG monitors. This study investigates machine learning approaches for classifying emotions based on ECG signals, utilizing K-nearest neighbors (KNN), support vector machines (SVM), and ensemble bagged trees (EBT) as classifiers. Three feature extraction methods are examined – statistical features in the time-domain, signal powers in different frequency bands from the wavelet transform, and scattering coefficients from the wavelet scattering transform. The study evaluates model performance across three diverse databases: YAAD, DECAF, and AMIGOS. Computer simulation results indicate that the approach based on wavelet scattering transform and ensemble bagged trees achieves the highest classification accuracy of 84.6%, 82.3%, and 87.8%, respectively, for the above three databases. This work provides insights into optimal feature extraction methods for ECG-based emotion recognition, offering a comparative analysis of widely used machine learning models and signal transformation techniques