Gender in the eye of the beholder
05 August 2014 19:47 GMT

Gender classification has the power to aid traditional biometric systems by reducing the searchable dataset by half

Is it possible to look into a person’s eyes and tell whether they’re male or female. Apparently so reveals new research into soft biometrics – and with quite some certainty!

The scientists are about to reveal results that show the human iris can be used to predict human gender with an accuracy rate of more than 91%.

The research team, made up from scientists at the University of Chile and University of Notre Dame, will present their results at the inaugural International Workshop on Soft Biometrics, in Zurich, on September 7th, 2014, in conjunction with the European Conference on Computer Vision.

Soft biometrics do not reveal identity, but could, for example, be used in a biometric recognition framework, as accurate gender classification could help by requiring a search of only half of the subjects in the database.

Of course there are other uses for this sort of technology outside of biometrics – for example collecting demographic statistics for marketing.

To date most gender classification systems have focused on facial analysis.  The researchers say that the use of the iris is a “rather new topic”.

Earlier research has shown an accuracy of approximately 80% (and a similar percentage for determining ethnicity), but by using a new method for extracting information from the iris it now appears possible to raise this accuracy to 90% and beyond – which is competitive with the state of the art for this problem, the researchers say.

There is also room for improvement as these experiments only assessed one eye (the left eye). With modern scanners, the researchers say, which now capture both iris images it may be possible to fuse the results together and improve the overall accuracy.

The researchers paper will be entitled: Gender Classification from Iris Images using
Fusion of Uniform Local Binary Patterns.