Emotion Detection Using Voice

Abstract

The system developed intends to be used by individuals of different backgrounds such as medicine, forensic sciences, call centres, Judiciary. It requires interdisciplinary approach that involve concept of signal processing and AI. Human voice has certain attributes such as pitch, timbre, loudness and vocal tones that in some way or the other effects the quality of sound directly. The quality of sound that is produced by a person’s speech illustrates unconscious state of mind. The system uses voice signal to extract features and classify them into emotions. The algorithms used for the classification are SVM, MLPclassifier and RNN. The system showed best accuracy with deep learning algorithm RNN. Therefore, the system emotion detection uses RNN to classify the emotion classes angry, sad, happy, calm, neutral, disgust and surprised.