Neural Networks for Emotion Classification

TL;DR

This paper presents a neural network approach for emotion classification that automatically optimizes parameters and uses Powell's method for backpropagation, achieving 77% accuracy on the Cohn-Kanade database.

Contribution

It introduces an automatic parameter selection strategy and a novel backpropagation method using Powell's algorithm for emotion recognition.

Findings

Achieved 77% accuracy on Cohn-Kanade database

Developed an automatic neural network parameter tuning method

Demonstrated successful emotion recognition with neural networks

Abstract

It is argued that for the computer to be able to interact with humans, it needs to have the communication skills of humans. One of these skills is the ability to understand the emotional state of the person. This thesis describes a neural network-based approach for emotion classification. We learn a classifier that can recognize six basic emotions with an average accuracy of 77% over the Cohn-Kanade database. The novelty of this work is that instead of empirically selecting the parameters of the neural network, i.e. the learning rate, activation function parameter, momentum number, the number of nodes in one layer, etc. we developed a strategy that can automatically select comparatively better combination of these parameters. We also introduce another way to perform back propagation. Instead of using the partial differential of the error function, we use optimal algorithm; namely Powell's direction set to minimize the error function. We were also interested in construction an authentic emotion databases. This is a very important task because nowadays there is no such database available. Finally, we perform several experiments and show that our neural network approach can be successfully used for emotion recognition.