Prediction of Intensity Level of Pain using Computational Methods

Pain has contributed to most of the health problem throughout the world. Unfortunately, the value of medical interventions is self-effacing Thus, automatic pain detection has proved to be an emerging area of investigation. It opens a window for those patients who are helpless in rating their pain level or cannot verbally describe it. Feature extraction approaches such as ICA, PCA, LDA, Surf, and Sift are considered in order to improve the accuracy in terms of system performance. The research contributions constitute three aspects: (1) the pain image database reflected on the face of individuals suffering from pain is prepared, (2) extraction of some particular set of features using the image database of facial pain, and (3) finally, the experimental results and discussion of face pain recognition algorithms. Using two different databases, the experiments are conducted i.e., UNBC-McMaster shoulder Pain database and the other is on the self-prepared database implemented on a frame level and image level. The result represents 89.27% accuracy at frame level for detection of pain and 96.2% at the image level. The methodology achieves 87% accuracy for classifying frames in four different pain levels.