Neonate facial gender classification using PCA and fuzzy clustering
Hamid Hassanpour, Hossein Dehghan · 2010
This paper considers the problem of neonate gender classification using frontal facial image. Determining sex of neonates using facial image is a challenging issue even for human observers. We propose a new gender classification method for neonate facial image by employing Principal Component Analysis (PCA) and Fuzzy C-means Algorithm (FCM). In this approach, PCA is used to extract suitable features with reduced dimensional space. These features are then used to assign the image to an appropriate class, hence recognizing it as belonging to a boy or a girl. This technique can be used to assist physicians in recognizing intersex neonates. Compared to the clinical approaches, such as hormonal, genetic and radiological methods, the proposed approach is fast and inexpensive. In an experiment performed on 48 neonate facial images, the naive human observers could classify the gender with 58.33% accuracy while the proposed method outperformed with 91.66% accuracy.