Gender Classification Using Gradient Direction Pattern
This work addresses gender classification from facial images, which is an incremental improvement for applications like security or demographic analysis.
The paper tackles gender classification by introducing a novel Gradient Direction Pattern (GDP) feature descriptor that extracts local facial features from gray intensity differences and concatenates histograms from sub-regions into a single vector, achieving very high accuracy on the FERET database and outperforming traditional descriptors using a support vector machine.
A novel methodology for gender classification is presented in this paper. It extracts feature from local region of a face using gray color intensity difference. The facial area is divided into sub-regions and GDP histogram extracted from those regions are concatenated into a single vector to represent the face. The classification accuracy obtained by using support vector machine has outperformed all traditional feature descriptors for gender classification. It is evaluated on the images collected from FERET database and obtained very high accuracy.