Incorporating signature verification on handheld devices with user-dependent Hidden Markov Models

Proceedings of the International Conference on Frontiers in Hadwriting Recognition (ICFHR 2008)A dynamic signature verification system based on Hidden Markov Models is presented. For each user model, the number of states and Gaussian mixtures of the Hidden Markov Model is automatically set in order to optimize the verification performance. By introducing this userdependent structure in the statistical modeling of signatures, the system error rate is significantly decreased in the challenging scenario of dynamic signature verification on handheld devices. Experimental results are given on a subset of the recently acquired BIOSECURE multimodal database, using signatures captured with a PDAThis work has been supported by the Spanish Ministry of Education under project TEC2006-13141-C03-03