Voice Based Emotion Detection Using Convolutional Neural Network

Abstract

Voice signal emotion identification is a difficult but important task for many applications such as affective computing, mental health evaluation, and human-computer interaction. With the use of convolutional neural networks (CNNs) on audio signal spectrograms, this study suggests a unique method for emotion recognition. Effective emotion discrimination is made possible by the suggested CNN architecture, which is built to automatically extract pertinent information from the spectrograms. Experiments show that the suggested strategy reaches state-of-the-art performance in emotion recognition from voice signals. Our results point to the possibility of using CNNs for voice-based emotion detection tasks in a very efficient manner, which could lead to the development of more reliable and precise emotion recognition systems.