| Resumen |
Biomedical signals can be used to diagnose several affections of the human body. Nonetheless, they can also be used to describe a more general behavior of specific organs and how they respond according to the feelings and emotions of a person. Therefore, the YAAD dataset, which contains electrocardiogram (ECG) and Galvanic Skin Response (GSR) signals, is used in order to detect and classify seven different emotions from 25 different subjects. Stimulus is provoked to the subjects by exposing them to watch a collection of different videos that evoke emotions, such as anger, happiness, sadness, among others. Two different subsets are used in this research, a single-modal and multi-modal signals. In this work, we propose a series of preprocessing techniques to clean and resample the original signals, then a simple 1-dimensional convolutional neural network is implemented to perform the classification task. Moreover, two different types of validation methods were used to validate our results. We have achieved an accuracy over 95% for both validation methods on the multi-modal subset and an accuracy over 85% for the single-modal subset. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024. |
