Emotion Recognition from Speech using Spectrograms and Shallow Neural Networks

Emotions add meaning to the conversation between individuals and allow us to better understand each other. Besides, Human-Computer interaction has produced significant changes in recent years to satisfy the requirements and responsibilities of clients. From this angle, it would be perfect for machines to automatically recognize human feelings in order to improve communication and interaction both parts. Some of the works in this field are based on hand-crafted features and others are based on Deep Learning (DL) models. In this article, we will propose a SER (Speech Emotion Recognition) system in which we will combine the power of DL models in self pattern recognition together with the ability of working on small databases.