An approach to enhance age invariant face recognition performance based on gender classification

There is a need for facial biometric based authentication system that responds quickly despite large num-ber of registered users. The main challenge in these authentication systems is accuracy along with quick response time. The response time can be lowered by having lesser number of comparisons of the probe image with gallery images. To enhance the response time we propose gender classification as a pre-processing stage so that the validation of the user continues by taking only those facial images which are in accordance with the gender of the user. We have used Principal Component Analysis to create male and female Eigen spaces during the training phase of the classifier.To validate our classifier we have used ORL, Indian and FG-NET databases. Our proposed method gives improved recognition accuracy for the images having ageing variations.The proposed Gender classification model gives an improved accuracy of 57.44%, 59.13% and 61.13 % for the probe images taken from the FG-NET database with training images from ORL, Indian and FG-NET databases respectively as compared to PCA.Also the number of comparisons are reduced by an average of 38% by incorporating the gender classification as the pre-processing stage in the age invariant face recognition system.(c) 2021 The Authors. Published by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

In facial biometric-based authentication systems, gender classification can be used as a preprocessing stage to improve response time and recognition accuracy by using only facial images that are consistent with the user's gender.