Forgery detection by local correspondence

An approach for forgery detection of offline signatures using local correspondence is presented. A local correspondence is estabilished between a model and a questioned signature to solve the offline problem. The cost of the matching of consecutive stroke segments is determined by comparing a set of geometric properties of corresponding substrokes. The Gaussian statistical model is used to determine the threshold of decision making. The writer-dependent information embedded at the substroke level was examined to capture the unballistic motion and tremor information in stroke segment.