A Survey on Automatic Emotion Recognition Using Audio Big Data and Deep Learning Architectures

Abstract

Traditional speech emotion recognition is based on the pipeline of pre-processing, feature extraction, dimensionality reduction and classification. Recognition performance such as the accuracy largely depends on the professional feature engineering and the classifier, which will be more difficult in the scenario of big data. Recently, many emotion researchers trend the direction to automatic emotion recognition from the raw signal, the motivation behind this is that neural network can learn representation and find the final result automatically. This work focuses on categorization and reviews on the current progress on end-to-end speech emotion recognition problems. In this survey, we discuss the requirement of the network model, process procedures and current achiements. We also explore some potential future issues in speech emotion recognition.