Gender classification for digital signage solutions using facial images

Gender classification has found application in a myriad of fields such as surveillance, interaction between humans and computers, face recognition and most recently in digital signage for gender-targeted advertising. There are several existing methods of gender classification such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Support Vector Machines (SVM). All these methods have their own set of advantages and limitations. This research paper aims at studying the performances of afore mentioned methods in classifying gender of faces belonging to Indian ethnicity. The author also proposes using the method of Multiple Kernel Learning, where a kernel model is derived from a linear combination of weighted base kernels as opposed to the conventional method of choosing a single kernel and optimizing its parameters.