Human Age and Gender Estimation using Facial Image Processing

The increasing demand of smart security systems has enhanced the demand for the proper identification and verification of a person. In this context, accurate estimation of age as well as proper identification of gender is highly significant. Therefore, in this work, we have implemented two separate methods with satisfactory runtime and efficiency to estimate both human age and gender using facial images. Our image processing based method involves comparison of some features extracted from the post-processed facial images of people of various age ranges followed by some edge-detection procedures, creation of binary masks and evaluation of wrinkle densities. Afterwards, thresholds were set via Naïve Bayes Classification to estimate classes. For the assessment purpose, we developed a database namely BUET facial database, which consists of images of both male and female of diverse ages. For the developed database, our proposed algorithm exhibits 76.3% accuracy in the age group classification while it shows 86.6% accuracy in the gender classification. Apart from BUET facial database, our developed algorithm has also been tested in three other databases and compared its performance with the reported literature for these databases. The mean absolute error is almost below 5.0 for this work, whereas, others exceed 5.0 in most of the cases. Moreover, the proposed algorithm exhibits reasonably good accuracy under different lighting conditions of images as well. Our study would provide further insight into the choice of appropriate features for the efficient and accurate estimation of the age and the gender of a person.