A Novel Approach for Continuous Pain Intensity Estimation
In this study, a novel approach for continuous pain intensity estimation based on facial feature deformations is presented. The proposed approach is based on the fact that the shape and appearance of facial features get deformed due to pain. The shape def
- PDF / 270,405 Bytes
- 8 Pages / 439.37 x 666.142 pts Page_size
- 81 Downloads / 202 Views
Abstract In this study, a novel approach for continuous pain intensity estimation based on facial feature deformations is presented. The proposed approach is based on the fact that the shape and appearance of facial features get deformed due to pain. The shape deformation caused due to pain is computed using Thin Plate Spline (TPS). The non-rigid parameters are used as representative of facial feature deformations and affine transformation parameters are ignored. The deformation of appearance features is extracted using local binary pattern features. The shape and appearance features are fed to relevance vector regression separately and jointly for pain intensity estimation. The pain intensity estimation is carried directly (by estimating the pain intensity from facial feature deformation) and indirectly by first estimating the Action Unit intensity and then computing the pain intensity. For assessment of the proposed approach, we have selected the popularly accepted UNBC-McMaster Shoulder Pain Expression Archive Database. Experimental results ensure the efficacy of the proposed approach for pain intensity estimation.
Keywords Thin plate spline Local binary pattern Pain detection Pain intensity estimation Relevance vector regression Facial feature descriptors
1 Introduction The reliable measure of pain intensity is the one of the major concern of medical practitioners and has drawn much attention. Earlier this task was done by analyzing the reports generated by patients themselves or by taking their interviews. Later, Neeru Rathee (&) Maharaja Surajmal Institute of Technology, Guru Gobind Singh Indraprastha University, New Delhi, India e-mail: [email protected] Dinesh Ganotra IGDTUW, New Delhi, India e-mail: [email protected] © Springer Science+Business Media Singapore 2017 R. Singh and S. Choudhury (eds.), Proceeding of International Conference on Intelligent Communication, Control and Devices, Advances in Intelligent Systems and Computing 479, DOI 10.1007/978-981-10-1708-7_50
443
444
Neeru Rathee and Dinesh Ganotra
Visual Analog Scale (VAS) was used for evaluating pain intensity [16]. The above methods were very convenient and were being widely used as they did not use advance technology [2, 5, 9]. In late 1980s, when the development in the field of facial expression came into existence [14], then researchers focused on detection of pain from facial images. The pioneering work in this direction was done by measuring pain in terms of Facial Action Coding System (FACS) [18]. The fascinating feature of such systems is that they can be used in real time for taking care of the patients in intensive care unit [1, 12, 13]. Moreover, it was helpful to those who need continuous monitoring. The pain intensity estimation was a binary problem earlier and researchers achieved success diagnose whether pain is present or not [1, 8, 13]. Recently, researchers are working toward measuring the various levels of pain [7]. The above-mentioned methods represent facial features either by shape features
Data Loading...
