Gender Classification by Information Fusion of Hair and Face
Zheng Ji, Xiaochen Lian, Bao‐Liang Lu · 2009
We have presented a modified geometric hair model for extracting hair features for gender classification. By using this model, hair features are represented as length, area, color, splitlocation, and texture. In order to integrate the outputs of both hair classifier and face classifier that use hair features and face features, respectively, we have proposed a classifier fusion approach based on fuzzy integral theory. The experimental results on three popular face databases demonstrate the effectiveness of the modified geometric hair model and the proposed classifier fusion method. From the experimental results, we can obtain the following observations. a) Hair features play an important role in gender classification; b) Face features are more critical than hair features to gender classification; c) Implementing the fusion of hair and face classifiers can achieve the best classification accuracy in all of the cases; d) The proposed fusion method can improve the classification accuracy dramatically when the performance of all the single classifier is not good. From this study, we believe that more external features such as hair and clothes should be integrated into face features to develop more reliable and robust gender classification systems.