Research On The Extraction Of Neonatal Pain Facial Expression Based On Log-Gabor

Pain may bring uncomfortableness and have a bad effect on the growth of neonate in thefuture.Currently, neonatal pain assessment mainly depend on professional medical workers, it notonly needs time and energy, but also may result in the mistake. So, developing an objective anduserful tool of neonatal pain assessment is significant.The researched work of this paper is as follows:(1) This paper has researched serveral methods of feature extraction and presents the methodof feature extraction of neonatal pain expression based on Log-Gabor Wavelet.(2) This paper has researched the theory of PCA on dimension reduction, and some othermethods, like LDA and2DPCA. During the experiment, this paper uses PCA to reduce thedimension of feature and analyses the relationship between the dimension after reducing andrecognition rate.(3) This paper compares the performance of several methods of classification, including SVMwith linear kernel, SVM with a polynomial kernel(degree=2,3,4), SVM with radial basisfunction(RBF) and so on.(4) This paper compares and analyses the recognition result of Gabor and Log-Gabor methodsunder the same condition.(5) After learning the knowledge of GUI, this paper has developed a convenient and usefulgraphical user interface.The result shows: Using Log-Gabor to do feature extracting, using PCA to do feature reducing andusing SVM to do classifying can achieve a satisfied performance. The best recognition rate can reach89%.