Performance Enhancement For Gender Recognition Using Trainable Bank of Gabor Filters and NCA

Abstract

Classifying male and female images based on facial information is a significant application for security issues destined for some innovative evolving fields, such as retail advertising and marketing. In this research work, an image processing technique is used to segment facial images of GENDER-FERET dataset, and choose multi interest prototypes automatically using trainable bank of Gabor filters. Spatial pyramid algorithm with three levels is used to extract the features for each prototype. Enhancement is accomplished in two main steps. First, the descriptors are ranked and most significant ones are identified using Neighborhood Component Analysis feature selection algorithm. Second, Cubic Support Vector Machine (Cubic SVM) classifier is applied. Several classification performance metrics are measured such as accuracy, testing and training time that are recorded better results. A classification rate of around 95.8% was achieved by using the proposed model compared to 93.7% for the related works.