Research On Neonatal Pain Expression Recognition Based On LBP-top Feature

Researches have shown that to neonates, especially those premature or severe newborn who need to receive a lot of repeated painful operations, pain will cause negative impact. Because neonates can not use language to express themselves, neonatal pain assessment is conducted by professionals and has many limitations, so it is siginificent to develop a neonatal pain facial expression recognition system.The researched work of this paper is as follows.(1) This paper establishes the neonatal facial expression database based on 8000 frame images choosen in the 124 neonatal videos which are preprocessed, each kind has 2000 pieces.(2) In view of the particularity of the neonatal facial expression the Adaboost algorithm is proposed. On the basis of the original Haar features, the asymmetric Haar-like eatures are introduced. To avoid error or invalid detection, the L-K optical flow method based on the yramid model and Kalman filter is proposed for real-time tracking. In order to improve the speed and accuracy, the corner point is used for detection and the screening mechanism is optimized.(3) For the limitation of Logic Binary pattern(LBP) only considering the space texture, the LBP-top is used for feature extraction, which efficientiy represents the video expression features. The SVM parameters are optimized with the improved grid search method, and are obtained by cross validation, which better meets the real-time performance of the system.(4) After learning the knowledge of Visual Studio 2010 and MFC,the paper has developed a neonatal pain expression recognition system, and successfully implements two classification(calm, pain) and four classification, the user interface is friend.