A visual programming toolkit demonstrator for offline handwritten forms recognition

Successful commercialisation of off-line handwritten forms-based applications requires integration of numerous different handwriting recognition components with form-specific syntactic and contextual knowledge. To support these requirements, we are developing a handwritten forms recognition toolkit which is based upon a visual programming paradigm. In this paper, we outline the main components of the toolkit, and then describe an initial demonstrator for handwritten postcode recognition which has been implemented using it. The demonstrator currently achieves a postcode recognition rate of 63.7% on a sample of 748 testset images from the Essex handwritten address database, corresponding to a raw alphanumeric character recognition rate of 93.8%, whereas the actual raw alphanumeric recognition rate is only 75.5%. More importantly, the components utilised in the demonstrator conform to a standard specification which allows them to be readily re-used in other offline handwriting recognition applications.