Abstract
The advancements in neural networks and the on-demand need for accurate and near real-time Speech Emotion Recognition (SER) in human–computer interactions make it mandatory to compare available methods and databases in SER to achieve feasible solutions and a firmer understanding of this open-ended problem. The current study reviews deep learning approaches for SER with available datasets, followed by conventional machine learning techniques for speech emotion recognition. Ultimately, we present a multi-aspect comparison between practical neural network approaches in speech emotion recognition. The goal of this study is to provide a survey of the field of discrete speech emotion recognition.
Highlights
Speech emotion recognition is the task of recognizing emotions from speech signals; this is very important in advancing human–computer interaction: Human computer interaction is characterized as consisting of five major areas of study: research into interactional hardware and software, research into matching models, research at the task level, research into design, and research into organizational impact [1]
Before the extensive employment of deep learning, Speech Emotion Recognition (SER) was relying on methods like hidden Markov models (HMM), Gaussian mixture models (GMM), and support vector machines (SVM) along with lots of preprocessing and precise feature engineering [4,5,6]
In earlier efforts to recognize emotions from the speech signal, almost all the implementations were based on machine learning and signal processing methods; following the same path of automatic speech recognition, there were many implementations based on SVM, GMM, and HMMs
Summary
Understanding one’s feelings at the time of communication is constructive in comprehending the conversation and responding appropriately. Along with all major problems in machine learning, SER has started to gain an advantage from the tools made available by deep learning. Before the extensive employment of deep learning, SER was relying on methods like hidden Markov models (HMM), Gaussian mixture models (GMM), and support vector machines (SVM) along with lots of preprocessing and precise feature engineering [4,5,6]. With deep learning making up most of the new literature, the results are going up from around 70% accuracy to the upper 90s in controlled environments
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
