Towards Pain Recognition in Post-Operative Phases Using 3D-based Features From Video and Support Vector Machines

Abstract

In this article first results are presented of our new method for vision based recognition of pain from facial expressions. Incorporating a series of image processing, low level computer vision and pattern recognition techniques the proposed method not only detects facial expressions of pain but also ones related to emotion or neutrally speaking subjects. In order to extract the facial features through image processing our approach uses color and gradient information along with a head contour model. In the 3D related part a camera model is included along with an initial registration step, in which we use method that automatically builds a person specific face model from stereo. By using this model information 3D geometric feature measures are determined and normalized through the application of photogrammetric techniques. A support vector machine based classifier is trained and applied to these features. The presented approach leads to minimal mixing between different expression classes. Also, our framework achieves robust and clear classification results while reaching high performance.