2D vs 3D online writer identification: a comparative study

Abstract

Nowadays, different automatic systems for writer identification and verification are available. On-line writer identification through automatic analysis of handwriting acquired with a tablet has been widely studied. Furthermore, the recent development of Commercial Off-The-Shelf (COTS) wearables with integrated inertial measurement units (IMUs) recording limbs movement allows the study of handwriting movements executed on the air. The goal of this paper is to compare the performance of an online writer identification system while processing 2D data acquired by a tablet while writing on-paper and 3D data acquired by a smartwatch while writing on-air. To this end, a database of handwriting samples produced by the same writers while writing the same symbols in the two modalities has been built up. The results of the study show a performance gap smaller than 5% between the 2D and 3D top implementations of the system, confirming that 3D handwriting is a promising alternative for developing wearable user authentication system.