A Study of Emotion Classification of English Voice Data Using Machine Learning And Comparative Verification With Japanese Model

Abstract

In recent years, the number of those suffering from mental health problems has been increasing due to lifestyle changes caused by COVID-19. Against this background, this research will study the elemental technology for a system that supports the human mind, such as a system in which artificial intelligence analyzes emotions from voice and applies words of encouragement when it detects negative emotions, which is one of the efforts to solve the mental issues of modern people. In order to build a highly accurate voice emotion classification model, this research focuses on the handling of data, which are challenges in speech recognition. The authors have previously created emotion classification models using decision trees and neural networks, focusing on the relationship between words and emotions. In this paper, we also focus on data characteristics and use English voice data from the IEMOCAP database. Based on the trends in models that have been constructed using Japanese voice data, the spoken emotion classification in English will be carried out. The trends in forumant. Compare the accuracy and trends of the models in different languages, and use this information to improve the accuracy of the emotion classification models.