Audio Emotion Detection - using Convolutional Neural Networks

Abstract

In today's society, one of the most crucial marketing strategies is emotion detection. For a person, you might tailor several things in order to best fit their interests. Due to this, the decision is made on the research work, where a person's emotions could be identified based just on their speech, allowing us to handle a variety of AI-related applications. The proposed Model gives 75% accuracy. To create any successful model, you need to follow some steps. Steps such as first of all you have to do data preprocessing, after there is a step such as data visualization because audio is cannot be seen through human eyes. Then there is a step named as data augmentation in which dataset quality is improved by adding some features. After doing overall data preparation, step such as feature extraction with the spectral features such as MFCC, and TEO is done. For model training and testing purposes, CNNs algorithm is used.