Construction of an Emotion Estimation Model Using EEG and Heart Rate Variability Indices as Features by Machine Learning

Abstract

Several human emotion estimation which apply machine learning have been under studying. These studies are expected to be applied to healthcare and medical diagnosis. Recently, EEG and heart rate variability indices were used for constructing emotion estimation model. The model accuracy was 0.80 in classifying the four emotions of joy, anger, sorrow, and pleasure. However, for applications in health care and medical diagnosis, the model with an accuracy of 0.80 may not be sufficient. Therefore, for the purpose of improving accuracy, this study extracted and select feature of EEG and heart rate variability indices. Then, the emotion estimation model was constructed by deep learning. As a result, the accuracy was 0.98 in this study while it was 0.50 in the features used by the previous study. Therefore, it was confirmed that the accuracy was improved.