Methodological Approach for Machine based Expression and Gender Classification

Machine based Gender classification is one of the challenging problem to the computer science researchers. The effortless ability exhibited by a two year kid for the same needs immense computation power for the computing machines. Many people attempted this problem by using different psychological characteristics such as handwriting, speech recognition, query response etc. These computational intensive techniques are application based and classification efficiency is limited. Some of the face identification systems have used the pixel based information, Eigen faces and geometrical relations of the facial features. We present a neural network-based upright invariant frontal face detection system which can classify the Gender based on the facial information. In our approach we club the pixel based and geometric facial features to increase the reliability of classification process. The use of pi-sigma neural network and the Cyclic Shift Invariance Technique enhances the robustness of classification process.