Predictability and correlation in human metrology

Human metrology provides an important soft bio-metric, which can be used in challenging situations such as human identification at a distance, when traditional biometric traits such as fingerprints or iris cannot be easily acquired. We study the problem of predictability and correlation in human metrology, using the tools of uncertainty and differential entropy. We show that while various metrological features are highly correlated with each other, there exists some correlation clusters in human metrology, whereby measurements in a cluster tend to be highly correlated with each other but not with the others. Based on these clusters, we propose a two-step approach for predicting unknown body measurements. Using the same framework, we study the problem of estimating other soft biometrics such as weight and gender.