Speech Emotion Recognition Using Machine Learning

Abstract

A recent research issue in the realm of Human Computer Interaction (HCI), Automatic Speech Emotion Recognition (SER) has a wide range of applications in a variety of situations. The goal of a voice emotion identification system is to automatically classify a speaker's utterances into one of five emotional states, including disgust, boredom, sadness, neutral, and happy, without the need for human intervention. The speech samples were taken from the Berlin emotional database, and the features extracted from these utterances included energy, pitch, linear prediction cepstrum coefficients (LPCC), Mel Frequency cepstrum coefficients (MFCC), Linear Prediction coefficients and Mel cepstrum coefficients (LPCMCC), and linear prediction coefficients and Mel cepstrum coefficients (LPCMCC). In order to categorise distinct emotional states, the Support Vector Machine (SVM) is employed as a classifier