Efficient Gender Classification Using Interlaced Derivative Pattern and Principal Component Analysis

Abstract

With the wealth of image data that is now becoming increasingly accessible through the advent of the world wide web and proliferation of cheap, high quality digital cameras it is becoming ever more desirable to be able to automatically classify Gender into appropriate category such that intelligent agents and other such intelligent software might make better informed decisions regarding them without a need for excessive human intervention. In this paper, we present a new technique which provides superior performance superior than existing gender classification techniques. We first detect the face portion using Voila Jones face detector and then Interlaced Derivative Pattern (IDP)extract discriminative facial features for gender which are passed through Principal Component Analysis (PCA) to eliminate redundant features and thus reduce dimension. Keeping in mind strengths of different classifiers three classifiers K-nearest neighbor, Support Vector Machine and Fisher Discriminant Analysis are combined, which minimizes the classification error rate. We have used Stanford University Medical students (SUMS) face database for our experiment. Comparing our results and performance with existing techniques our proposed method provides high accuracy rate and robustness to illumination change.