Early choke infant monitoring scheme

摘要

This paper come out with an infant behavior recognition scheme based on neural network. In this study, the infant face region is segmented based on the Principle Component Analysis. Two four of features, namely Mean, Variance, Skewness and Kurtosis are then calculated based on the information available from the infant face regions. Since each type of features in turn contains several different values, given a single fifteen-frame sequence, the correlation coefficients between those features of the same type can form the attribute vector of pain and normal facial expressions. Fifteen infant facial expression classes have been defined in this study. Support Vector Machine (SVM) corresponding to each type of those features has been constructed in order to classify these facial expressions. The experimental results show that the proposed method is robust and efficient. The properties of the different types of features have also been analyzed and discussed.