Real-Time Pain Detection Using Deep Convolutional Neural Network for Facial Expression and Motion

Abstract

At present, in every corner of the world, including developing and developed, countries got affected by infectious diseases such as the COVID-19 virus. Our objective was to create a real-time pain detection for everyone that can use it by themselves before going to the hospital. In this research, we used a dataset from the University of Northern British Columbia (UNBC) and the Japanese Female Facial Expression (JAFFE) as a training set. Furthermore, we used unseen data from webcam or video as a testing set. In our system, pain is divided into three categories: mild, moderate-to-severe-to-painful, and severe. The system’s efficiency was assessed by contrasting its results with those of a highly qualified physician. Classification accuracy rates were 96.71, 92.16, and 98.40% for the not hurting, getting uncomfortable, and painful categories. To summarize, our research has created a simple, cost-effective, and readily understood alternate method for the general public and healthcare professionals to screen for pain before admission.