Execução de testes de reconhecimento automático de dor em recém-nascidos usando imagens inteiras e marcações faciasi na base icope com o matlab®

Abstract

This work aims to perform automatic pain recognition tests in newborns, in which the parameter values of the method used were varied in the execution of the prototype in Matlab® aiming to increase the average success rates, thus contributing to the project "Research and application of facial recognition techniques", also considered 2 test files, one that uses feature extraction from facial markings and another of whole images, these images are of newborns between 18 and 72 hours from an image base available on the Internet. Comparative tests were performed on the project prototype between different methods of feature extraction, dimensionality reduction and classification. The tests showed a high rate of accuracy in the automatic recognition of pain in newborns, allowing the future development of an application for cell phones focused on the medical area.