Pain Detection From Facial Expressions

Abstract

This paper presents a novel approach for pain prediction based on facial expressions using a new points landmarks of the face. Despite advancements in technology, the evaluation and management of pain, especially among individuals with intellectual disabilities, remain challenging due to issues in communication. Therefore, our proposed approach aims to address this issue by leveraging facial expressions as a reliable indicator of pain levels, circumventing the communication barriers often present in this population. Our method comprises four phases: i) Data collection and preprocessing ii)- Face detection and cropping iii)- Features extraction iv)- classification phase. By exploring different machine learning algorithms, including the Random Forest classifier, we obtain great performance compared to the existing works in the literature. The results obtained highlight the effectiveness of our approach in accurately predicting pain levels from facial expressions. In the case of four pain classes, our approach achieves an accuracy of $89 \%$, while in the case of three pain classes, it achieves an accuracy of $92 \%$. These high performances underline the reliability and accuracy of our approach for pain prediction.