Sentiment Analysis of Human Speech using Deep Learning

Sentiment is an integral part of human psychology, influencing our behavior and decision-making in various situations. One powerful tool for expressing emotions and sentiments is speech, which can convey a wealth of information about an individual’s state of mind. With the increasing demand for real-time Speech Emotion Recognition (SER) systems, this field has become a popular area of study and research. The objective of this research paper is to build an effective SER system that can accurately recognize and classify emotions from speech. We have utilized Bi-directional Long Short-Term Memory (Bi-LSTM) and Convolutional Neural Networks (CNN) model for recognizing and classifying emotions and considered various acoustic features to enhance the accuracy of this classification. In this paper, we use the publicly available Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) and Crowd Sourced Emotional Multimodal Actors Dataset (CREMA-D) datasets for the training and testing of our model. The results of our research show that the proposed SER system has achieved a high accuracy of emotion recognition and classification, which demonstrates the effectiveness of using deep learning models and acoustic features in the process.