Majority voting for improved signature verification

Abstract

The rich and varied nature of the information with which to characterise signatures gives rise to a number of issues which must be addressed for a viable signature verification system to be implemented. However, signature verification solutions do not generally make effective use of these multiple sources of information and may miss the opportunity to implement a sophisticated signature verification system utilising a diverse range of sources of description for signatures. In order to achieve a successful combination of information sources and types, the problem of combining disparate forms of signature information, including both continuous and discrete features, and parametric and signal-based methods of signature representation must be addressed. The utilisation of a number of different sources of identifying information is commonly described in terms of multiple experts, where each individual expert is defined as a single set of features or classifier type and the overall system is formed by a combination of these experts. The combination of different sources of information about a signature, in the form of different feature sets and classification methods, provides an opportunity to develop an improved level of verification compared to the use of a single set of descriptors. In order to investigate this multiple expert architecture empirically, eleven experts are considered, formed from both static and dynamic signature data, with the intention of using complementary sources of information about a signature. (4 pages)