A Review on Deep Learning Algorithms for Speech and Facial Emotion Recognition

Charlyn Pushpa Latha,Mohana Priya

doi:10.11591/aptikom.j.csit.118

Abstract

Deep Learning is the recent machine learning technique that tries to model high level abstractions in data by using multiple processing layers with complex structures. It is also known as deep structured learning, hierarchical learning or deep machine learning. The term “deep learning" indicates the method used in training multi-layered neural networks. Deep Learning technique has obtained remarkable success in the field of face recognition with 97.5% accuracy. Facial Electromyogram (FEMG) signals are used to detect the different emotions of humans. Some of the deep learning techniques discussed in this paper are Deep Boltzmann Machine (DBM), Deep Belief Networks (DBN), Convolutional Neural Networks (CNN) and Stacked Auto Encoders respectively. This paper focuses on the review of some of the deep learning techniques used by various researchers which paved the way to improve the classification accuracy of the FEMG signals as well as the speech signals.

Highlights

Emotions are experienced from an individual point of understanding
The algorithmic power of deep learning can be merged with the efficiency of neuromorphic processors, bringing the promise of embedded, intelligent, brain-inspired computing one Experimental results show that cascading finetuning approach achieves better results, compared to a single stage fine tuning with the combined datasets with an overall accuracy of 48.5% obtained in the validation set and 55.6% in the test set, which compares favorably to the respective 35.96% and 39.13% of the challenge baseline Two deep learning models namely Deep Belief Networks (DBN) and Convolutional Neural Networks (CNN) models were proposed and the efficiency obtained were 65.22% and 95.71% respectively for both person independent and person dependent approaches
A solution is proposed to the problem of „contextaware‟ emotional relevant feature extraction, by combining Convolutional Neural Networks (CNNs) with Long Short Term memory (LSTM) networks, in order to automatically learn the best representation of the speech signal directly from the raw time representation A complete system for the 2015 Emotion Recognition in the Wild (EmotiW) Challenge is proposed

Summary

Introduction

Emotions are experienced from an individual point of understanding. It is mostly related with mood, temperament, personality, and disposition. Ankit Awasthi [3] used the combined algorithms namely Restricted Boltzmann Machine and the Deep Belief Networks and developed a model that attracted the important features learnt by the model by changing the number of hidden layers and units.

Results

Conclusion