Classification of Human Emotion from Speech Data Using Deep Learning

Emotions are complicated mental processes that react to stimuli in the environment. It is a process that allows people to alter and express their feelings in a variety of settings. Humans can express their emotions in a variety of ways depending on the occasion. Therefore, it is difficult to detect and comprehend true emotions. Predicting the emotions of the interlocutor aids in determining the best course of action in specific situations, such as treating patients with depression or those who require psychotherapy. This study developed deep learning models to classify humans' emotions by using their speeches. Then, humans' voices were classified based on five emotional categories, including normal emotion, anger, surprise, happiness, and sadness. The objectives of this study were to 1) compare the performances of two classifying models, i.e., Convolution Neuron Networks (CNN) and Long Short-Term Memory (LSTM), and 2) propose the most appropriate model for classifying humans' emotions from speech recognition. It revealed that classification results generated by LSTM outperformed CNN. The four-layer LSTM was found to be capable of recognizing human speech.