Emotional Analysis using Spiking Neural Networks

Abstract

With Artificial Intelligence gaining popularity in every field. Emotions is the one thing that Artificial Intelligence or AI cannot replicate yet. One of the key reasons being that emotions are often complex and hard to understand even by humans who used to them. Not only are these emotions complex and hard to understand, humans often express their emotions in different ways. Some don’t even express it at all. While some, express it using big motions and exaggerated facial expressions. There’re also cases where it is expressed in the voice via small changes in the pitch and frequency of the words as well as the intonation. All of these different ways of expression as well as the complications of these emotions make it hard for it understood much less replicated. Emotional analysis is method to understanding emotions that are expressed by humans in various different forms. Emotional analysis is a popular problem statement, and is constantly worked on with the advancing technology. One such new technology is Spiking Neural Networks, a newer model of Neural Networks that is based on the biological spiking of neurons to pass information in the brain. In this paper, we propose the method of using SNN’s on multi-modal data for the purpose of emotional analysis. The multi-modal data that we used encompasses both physiological and physical signs of emotion. We have also tested uni-modals, bi-modals and multi modals of the same data.