The Research And Implementation Of Gender Identification System Based On The Whole Face And Facial Features

The biological characteristics of facial features are innate. They are unique and cannot becopied. Moreover, compared to other biological features, facial features are friendly. They are easyto follow. Face recognition has been applied to the security monitoring, identification,human-computer interaction and so on. People can identify the identity of target, gender, age, race,etc through facial images. In recent years, people have paid more and more attention, and graduallybecomes the focus of research.The main purpose of this paper is to design a certain reliability and real-time genderrecognition system. One of the main problems in gender recognition is the recognition accuracy isnot high in poor lighting conditions. In view of the above problems, this paper puts forward thesystem design method as follows: First of all, image preprocess is completed to detect face for oneframe. At the same time, detect the different face sub-regions. Then extract the whole face imagefeature using improved local binary patterns. In order to balance accuracy and speed, gray valuefeature is used as input for face sub-regions. And different classifiers are trained for whole face andits sub-regions, then gender identification result is obtained by the fusion of these differentclassifiers. Experimental results show that this method can solve the problem of adverse effects oflight in a certain extent. And the complementary of different facial sub-region improve the accuracyof gender identification.Finally, this paper introduces the implementation of a real-time gender recognition systembased on OpenCV library. The testing results show this system owns high performance both onaccuracy and time cost.