Emotion Recognition Based on Convolutional Neural Network (CNN)

Abstract

Humans consistently have the inborn capacity to perceive and recognize faces and emotions. Computers can now accomplish the same thing, which creates lots of new challenges in daily life. For example, emotion recognition can increase security, enable financial transactions without the need for real cards, and enable the identification of criminals and specialized treatment, among other things. Therefore, face detection and emotion recognition is a prominent and futuristic research topic. In the near future, open-source projects will get more importance rather than licensed ones. In this connection, it is proposed to utilize a python library for face detection and recognition. Kaggle web resources are providing open-source Face Emotion Recognition (FER) datasets. With the help of datasets, there are seven types of emotions: happy, sad, fear, disgust, angry, neutral, and surprise. It is proposed to use image augmentation to improve emotion recognition by building a six-layered Convolution Neural Network (CNN) in Python using the Keras toolkit.