Speech Emotion Recognition System using Librosa for Better Customer Experience

Abstract

Call center employees usually depend on instinct to judge a potential customer and how to pitch to them. In this paper, we pitch a more effective way for call center employees to generate more leads and engagement to generate higher revenue by analyzing the speech of the target customer by using machine learning practices and depending on data to make data-driven decisions rather than intuition. Speech Emotion Recognition otherwise known as SER is the demonstration of aspiring to perceive human inclination along with the behavior. Normally voice reflects basic feeling through tone and pitch. According to human behavior, many creatures other than human beings are also synced themselves. In this paper, we have used a python-based library named Librosa for examining music tones and sounds or speeches. In this regard, various libraries are being assembled to build a detection model utilizing an MLP (Multilayer Perceptron) classifier. The classifier will train to perceive feeling from multiple sound records. The whole implementation will be based on an existing Kaggle dataset for speech recognition. The training set will be treated to train the perceptron whereas the test set will showcase the accuracy of the model.