Research On Neonatal Pain Expression Recognition From Video Sequence

Neonates have pain perception but they can not verbalize this feeling by themselves. Neonatal pain assessment is considered one of the most challenging tasks in neonatology. At present, neonatal pain assessment usually depend on the observations of health professionals, which may be affected by subjective factors more or less. Therefore, it is of great importance to develop an objective recognition system to help evaluating neonatal pain timely and effectively.In this thesis, we do further research on the neonatal facial expression recognition from video sequence and design a corresponding system to display the process and result. The research work of this thesis is as follows.(1) In order to determine the location of face, a face detection aigorithm based on skin color and Haar feature is proposed. Experimental results show that the method based on skin color can reduce the face detection range and reduce the detection time effectively. Facial position can be detected when the face area of the newborn is complete and the shooting angle is positive.(2) To meet the real-time demand of video sequence processing, the tracking algorithm based on pyramidal Lucas Kanade optical flow is used to track the movement of neonatal facial region. Meanwhile, the combination of forward and backward tracking is proposed to make the tracking result more accurate. Experiments show that this method can achieve good tracking effect even when the face is under partial occlusion and there is angel offset.(3) LBP and SVM are used to extract and recognize the feature of neonatal pain expression respectively. Uniform LBP patterns and rotation invariant LBP patterns are combined to improve operational speed and the effect of image block is taken into account in order to get more details of facial expression. To solve the parameter selection problem in SVM, grid search method is used to find the best penalty factor and kernel parameter for the recognition system.(4) MFC and OpenCV are used to build the overall framework of the system and obtain the image processing algorithm respectively. Whole system running result shows that this design for recognizing neonatal pain expression in video sequence can achieve basic facial expression recognition function and has the advantages of friendly interface, simple and easy to use.