Influence of Emotions on Prosody and Epoch in Assamese Speech
Abstract
With the advancement in human-computer interaction, the most growing and rising research area in the realm of human-machine interaction and speech processing is automatic emotion detection in speech. Finding the necessary speech traits that separate the emotional content of a speech is a time-consuming and difficult endeavor. The goal of this research work is to comprehend and analyze the function of emotions in prosody and excitation source characteristics (Epoch) of a speech signal to create a robust Assamese Speech emotion recognition system. Finally, we attempted to distinguish emotions by incorporating the derived relevant emotion-specific features into Machine learning techniques such as Multinomial Logistic Regression and Neural Network, and observed that Multinomial Logistic Regression has a higher classification accuracy than Neural Network. This study strategy will encourage readers to consider prosody, emotion, and speech from a variety of perspectives, as well as contribute to the development of a robust Assamese Speech Emotion Recognition system for real-time applications.