Computerised gender classification (part 1: eyebrows)

Can you tell a person’s gender from just a glimpse of their eyebrows? Could a computerised system do the same? To find out, a project was undertaken by Yujie Dong (of the Holcombe Department of Electrical and Computer Engineering, Clemson University) and Damon L. Woodard (now at the Biometrics and Pattern Recognition Lab, University of Florida) in 2011. The team used Minimum Distance Classifier (MD), Linear Discriminant Analysis Classifier (LDA) and Support Vector Machine Classifier (SVM) algorithms to extract eight eyebrow features from images (rectangularity, eccentricity, and isoperimetric quotient etc.) and were then able to determine a subject’s gender with an average accuracy of 96%

Eyebrow-Gender-DetectionBecause of their comparatively robust demarcation, eyebrows, say the team, provide a more reliable method of gender classification when compared to other ‘periocular’ [around the eye] features such as skin wrinkles – especially when the images are non-ideal, or, as some describe them ‘in-the-wild’. See: ‘Eyebrow shape-based features for biometric recognition and gender classification: A feasibility study’  (in: Biometrics (IJCB), 2011 International Joint Conference on. IEEE, 2011, pp. 1–8.)

Coming soon : Computerised gender classification (part 2)