Signature verification system for Point of Sale terminals

Abstract

Most people are used to signing documents and because of this; it is a trusted and natural method for user identity verification.

This dissertation proposes a statistical model for automatic signature verification. This model recognises that repeated signatures by the ‘owner’ are similar but not identical. Our model consists of a template signature for each individual, and several factors, which allow for variations in each rendition of this template. These variations include the speed of writing as well as slowly varying features such as size, number of strokes, number of points in signature and direction of start. The estimated template represents the mean of sample signatures from an individual, and the variations in the factors can be used to establish several measures of variance. These quantitative measures are essential for reliable signature verification. The novel idea in this model is that the weights associated with each of the features are dynamically varied for each user. That is, weight associated with feature ‘size’ will be higher for a user who signs in consistent size most of the time than the other users.

Going one step further this dissertation also proposes a system, which can be incorporated with the Point of sale (POS) terminals at the retail merchants. Primary objective of the system is to reduce signature related fraud and gain customer confidence in eBussiness.