Emotion Recognition using Machine Learning and ECG signals

Abstract

Various emotions can produce variations in electrocardiograph (ECG) signals, distinct emotions can be distinguished by different changes in ECG signals. This study is about emotion recognition using ECG signals. Data for four emotions, namely happy, exciting, calm, and tense, is gathered. The raw data is then de-noised with a finite impulse filter. We use the Discrete Cosine Transform (DCT) to extract characteristics from the obtained data to increase the accuracy of emotion recognition. The classifiers Support Vector Machine (SVM), Random Forest, and K-NN are explored. To find the optimal parameters for the SVM classifier, the Particle Swarm Optimization (PSO) technique is used. The results of the comparison of these classification methods demonstrate that the SVM approach has a greater accuracy in emotion recognition, which may be applied in practice.

Applying Self-Supervised Representation Learning for Emotion Recognition Using Physiological Signals

“…The expression of emotions through physiological responses is a natural process, usually unconscious and controlled by the central nervous system, which makes it difficult for the subject to fake or mask his/her emotional reactions. Thus, the inference of emotions through physiological signals has advantages compared to inference from subjective experiences or behavioral responses [ 36 , 37 ].…”

“…Electrocardiography (ECG) is a record of the electrical activity generated by the heart during a time interval [ 45 ]. In the health field, it is an effective and non-invasive tool, which, in addition to providing data to diagnose abnormalities present in the heart, can also be used to identify the emotional states of individuals [ 46 ], since emotions can produce variations in the signals of the ECG [ 37 ].…”