Prediction of Human Emotions by Neural Oscillations

Abstract

Brainwave, better known as neural oscillation is generally a neural oscillation or an electric impulse which is repetitive, often referred to as a rhythmic activity formed due to the interaction between various neurons in the CNS (Central Nervous System). All the neurons sync with the help of pacemaker cells or through the ability of the neurons to quickly sync up which is also referred to as entrainment. Brainwaves can be read using EEG method. Electroencephalography, or EEG, is a method used to measure neural oscillations within the brain. In this method, certain electrodes are placed on the patient’s scalp to note the data regarding electrical functioning of neurons in cerebral cortex. EEG identifies the impulses or waves created during the time of a billion neurons being active all together and it also notes the signals from specific places around each electrode. It basically provides a diagram or a graph of electrical activity in the brain represented as waves having different frequency, shape and amplitude. EEG is particularly used to measure brain activity during a particular event like losing a competition, accomplishing something or even feeling sleepy. These types of brain activities are called event-related potential. In this paper we will be predicting emotional sentiments using various machine learning algorithms. We have performed statistical extraction of brainwaves to create a larger dataset that is then reduced to a much smaller dataset by feature selection method for experimentation. In general, we have focused on three sentiments – Positive, negative and neutral. Algorithms like–Logistic Regression (with/without PCA-Principal Component Analysis), Support Vector Machine (SVM), Random Forest, XG Boost have been used and then their predictions are used to get to a conclusion.