Deriving a subspace model for personal authentication in biometric - Dynamic signature case

This paper deals with personal authentication by biometrics and specifically by signature verification. In biometrics most common case is to register a single data set for authentication. That is why it is frequently not possible to apply popular pattern recognition methods which require a significant amount of positive and negative training data. We propose a new recognition method that uses a subspace derived from a set of test experiments data. By applying a subspace centred on the point designated as the registered data we can authenticate signatures of registered users.