Speech Emotion Recognition using Librosa Library

Abstract

All living human being on earth express their opinion through a language. Some meaningful voice makes a language. In all the call center giving service faces many problem when they talk to customer. Sometimes the employee do not understand the emotion that the customer express. From the past two decades speech emotion recognition from speech becomes an interesting topics to researcher’s. This study is about various deep learning based algorithms RELU and SOFTMAX as well as models called Long-Short Term Model (LSTM) for the purpose of recognizing the emotion from speech. The dataset (TESS) is consist of 5600 voice data which is divided in seven categories such as (fear, angry, disgust, neutral, sad, pleasant-surprise, happy). The LSTM gives accuracy of 99.20% and it is the highest accuracy.