LBP and Iris Features based Human Gender Classification using radial Support Vector Machine

Identification of sex plays a vital role in forensic and medico legal investigations. Redial kernel SVM base classifier is used for gender identification in this work and Iris crypt densities,Histogram of Oriented Gradients (HOG) and Local Binary Patterns (LBP)and are considered as the features for classification. The thesis conducted on 200 subjects (100 males and 100 females) in the age group of 18–60 years. Along with the Crypt count this work uses Histogram of Oriented Gradients (HOG) features for detection of orientation of human face. On basis of only HOG features we can only recognize the orientation and we get 67.85% accuracy of gender classification. Local Binary Patterns (LBP) found different in male and female face; hence this work uses LBP as another feature for classification, we get 80.55% classification rate when only LBP features are used. Iris Crypt densities on the right- and left-iris were determined using a newly designed layout and analyzed statistically, the proposed work results showed that females tend to have a higher iris-crypt density in both the areas examined, individually and combined. Differences in the crypt density can be used as an important tool for the determination of gender in cases where partial eye-iris are encountered as evidence. On basis of Crypt densities, we get 90% accuracy of gender classification. This work merged all three features and found 98.5% gender classification rate with Redial kernel Support Vector Machine (SVM) classifier. The work is done on MATLAB 2018b version and standard human face database is FERET for identification.