AN EMOTION-BASED NEURAL NETWORK

ABSTRACT

The goal of this experiment was to create a neural network that would be able to identify emotions from electroencephalogram (EEG) recordings. Movie clips from databases of emotion-eliciting stimuli were used in order to get subjects to feel the basic emotions of anger, disgust, fear, joy, and sadness. A minimum of ten three second EEG recordings were taken per subject per emotion. Using Matlab and the EEGlab extension, the data was pre-processed and divided into five clusters based on emotion, and then a neural network was trained on the recordings. The neural network was used to establish patterns in the data that could be used to label future EEG recordings from the dataset based on the emotion being experienced while the EEG recording was taken. After training for 1,000 iterations, a neural network that could identify EEG recordings of emotions from the subjects was created. Developments such as this could be used to assist people who cannot communicate in expressing their emotions, such as those who have had strokes, are mute, or have other conditions. This would assist professionals such as doctors or psychologists who need to understand what their patients or clients are feeling despite a communication barrier.