A Comparative Study of Deep Learning and Machine Learning Approaches in Speech Emotion and Gender Recognition System

Abstract

Emotions have a major impact on mental health and wellbeing of a person. This study focuses on the effective role that machine learning and deep learning approaches have in the early detection process of depression, thus preventing people from taking drastic measures. This can be done with the help of a speech emotion recognition system, through which one can identify and understand the emotional state of a person just by listening to them when they talk. In this work, audio datasets of Kannada and English languages are collected and classified into four categories: happiness, anger, neutral and sadness. For speech emotion recognition systems, the performance of Machine Learning algorithms is compared to deep learning algorithms when applied on both English and Kannada language datasets. Deep Learning algorithms show better accuracy. For speech gender recognition systems, a gaussian mixture model is used which gives satisfactory values for accuracy, precision and recall.