Human identification from body shape

We investigate the utility of static anthropometric distances as a biometric for human identification. The 3D landmark data from the CAESAR database is used to form a simple biometric consisting of distances between fixed rigidly connected body locations. This biometric is overt, and invariant to view and body posture. We use this to quantify the asymmetry of human bodies, and to characterize the interpersonal and intrapersonal distance distributions. The former is computed directly and the latter by adding zero-mean gaussian noise to the landmark points. This simulation framework is applicable to arbitrary shape based biometrics. We use gross body proportions information to model a computer vision recognition system.