Automatic analysis of handwritten documents

Citation

Rudall, B.H. (1999), "Automatic analysis of handwritten documents", Kybernetes, Vol. 28 No. 5. https://doi.org/10.1108/k.1999.06728eaa.005

Automatic analysis of handwritten documents

Keywords Automation, Cybernetics, Research, Technological developments

Abstract Reports and surveys are given of selected current research and development in systems and cybernetics. They include: Language interface, Automated automobile, Innovative space technology, Software reliability and safety, Automatic analysis of handwritten documents, High-tech musical instruments, Biological motors, Interplay between smell and the mind, Cybernetics and automation.

Automatic analysis of handwritten documents

Methods of analysing handwritten signatures have taxed researchers for many years. The ultimate goal of producing an algorithm that is capable of analysing a signature automatically to a high degree of accuracy still beckons. The potential applications are numerous and financially attractive to research groups. It has to some extent become the "Holy Grail" for computer scientists.

A further contribution to these endeavours is being made by researchers at the Electronic Engineering Laboratory at the University of Kent at Canterbury, UK. Their aim is to develop powerful tools to enable handwritten signatures to be automatically analysed

Researchers at the laboratory work on what is called their "signature verification" project which is focused on applying techniques to increase the reliability with which handwritten signatures can be analysed and automatically authenticated.

The principal investigator of the project, Professor Michael Fairhurst, believes that although this is the main aim it also has a broader long-term potential. He says that:

An exciting aspect of our work has been the emergence of techniques that are likely to have a very important generic element relevant to a range of problems arising from the processing of handwritten documents, in addition to increasing significantly the diversity of practical solutions where the signature verification can be used effectively.

The researchers at the University of Kent are exploring these techniques in collaboration with PenOp which is a software company which develops and markets signature processsing technology for electronic commerce applications. In order to help refine signature analyses extensive data collection trials have been undertaken in various public sites including some well-known high street stores. The researchers say that:

Although the main thrust of our work has been on investigating novel approaches to enhancing the robustness and reliability of automatic signature verification, we have also built on techniques previously established in our group.

The group has explored various avenues to improving signature verification. One of the most effective has been the development of better models to characterise an individual's signature. This has identified a number of techniques for "validating" a signature model to increase confidence in its power to represent genuine samples.

Another way of optimising decisions about authenticity which has been studied is the selective exploitation of information that characterises an individual's signing patterns and styles. Perhaps of greatest interest is the approach that is currently attracting most attention. According to Professor Fairhurst this has been one of trying to gain a better understanding of the process whereby forged imitations of a signature are generally attempted. This technique has been explored in public trials, we are told, that have been supported by the collaborating company PenOp.

It is interesting to note how this technique is approached. In a trial participants are asked to "imitate" a sample signature under a variety of commonly occurring situations, such as when a genuine signature is available, as with a stolen credit card, or when there is no such signature, for example when the forger has possession only of a cheque book. Professor Fairhurst says that: "It is important to understand how different conditions influence the sort of models which might be appropriate for maximising security protection".

The group believe that looking to the future development of efficient techniques for integrating different sources of authenticating evidence is one of the most potentially fruitful areas. They also observe that in addition to exploiting as much signature information as possible from different types of feature extraction processes, improved integration could draw more effectively on different, but complimentary biometric measures such as speech and facial characteristics.

Most cyberneticians and systemists would agree that the "spin-off" from such detailed studies will benefit many areas of application. It is fascinating to note that so many studies that start by working on one task rapidly produce techniques and methodologies that are likely to impact on a whole range of problems.

Further information about these studies and the programme that has been funded by the Engineering and Physical Science Research Council (EPSRC) can be obtained from the web links: http://www.eleceng.ukc.ac.uk/research/ (University of Kent, UK); www.penop.com (PenOp, software company).