A Review on Speech Emotion Recognition Using Machine Learning

Abstract

This paper focuses on the development  of a robust speech emotion recognition system using a  combination of different speech features with feature  optimization techniques and speech de-noising technique  to acquire improved emotion classification accuracy,  decreasing the system complexity and obtain noise  robustness. Additionally, we create original methods for  SER to merge features. We employ feature optimization methods that are based on the feature transformation and  feature selection machine learning techniques in order to  build SER. The following is a list of the upcoming events.  A neural network can use either of these two techniques.  As more feelings are taken into account, the feature  fusion-acquired SER accuracy falls short of expectations,  and the plague of dimensionality starts to spread due to  the addition of speech features, which makes the SER  system work harder to complete its task. This is due to the SER system becoming more complicated when voice  elements are added. Therefore, it is crucial to create a  SER system that is more trustworthy, has the most  practical features, and uses the least amount of computing  power possible. By using strategies that maximize current  features, it is possible to streamline the feature selection  process by reducing the total number of accessible  choices to a more reasonable level. This piece employs a  method known as Semi-Non Negative Matrix  Factorization to lessen the amount of processing trash that  the SER system generates. (Semi-NMF). This approach  can be used to change traits that are capable of learning  on their own.