A Novel Approach for Pain Intensity Detection by KAZE Features Ashutosh Vaish and Sagar Gupta Abstract Successful pain evaluation is one of the most difficult problems of clinical practitioners. Continuous pain evaluation of the patient in the intensive care unit (ICU) is burdensome and expensive. An individual, generally a nurse, has to be present at all times to gage the level of pain of a person during as well as after an operation. This research paper is an attempt to automate the pain intensity detection through the scrutiny of the facial features. The intent of this research is to provide a new method of detecting the intensity of pain, i.e., by Kaze features. Kaze features differ from the previous features in the sense that the blurring made by it is locally adaptive. The building of Fisher vector is done through GMM. This paper also reviews the various features used for the purpose of pain detection. The high accuracy of 91.8% instills the confidence in using of Kaze for facial features and makes the realization of an actual state of the art closer than before. Keywords Real time · Pain intensity detection · Facial expressions · Kaze features 1 Introduction Managing and assessing pain is a difficult task. The most widely used methods for assessing pain is the patient self-report as no special skill set is needed and is convenient. Other methods which have been successfully used for pain assessment include a visual analog scale and clinical interviews [1]. The scale varies from zero pain (0) implying no pain to high pain, i.e., the worst imaginable pain (100) on a scale of 100 mm [2, 3]. A. Vaish IIIT Delhi, New Delhi, India e-mail: ashutoshv@iiitd.ac.in S. Gupta (B ) IIITD, Okhla Industrial Area, New Delhi, India e-mail: absagargupta@gmail.com; sagar18174@iiitd.ac.in © Springer Nature Singapore Pte Ltd. 2019 V. Nath and J. K. Mandal (eds.), Proceedings of the Third International Conference on Microelectronics, Computing and Communication Systems, Lecture Notes in Electrical Engineering 556, https://doi.org/10.1007/978-981-13-7091-5_1 1