Improved age prediction from biometric data using multimodal configurations

The prediction of individual characteristics from biometric data which falls short of full identity prediction is nevertheless a valuable capability in many practical applications. This paper considers age prediction in two biometric modalities (iris and handwritten signature) and explores how different feature types and classification strategies can be used to overcome possible constraints in different data capture scenarios. Importantly, the paper also explores for the first time the use of multimodal combination of these two modalities in an age prediction task.