The on-line isolated character recognition problem is in many respects a mature research area with several in- dustrial applications. When it comes to on-line cursive word recognition progress has been much slower. One of the obstacles refraining simple extension of the val- idated single character recognition methods to cursive word recognition is the fact that most methods rely on various forms of global normalization techniques which makes it difficult to define a fair way of adding such match- ing distances for sequences. This paper proposes a new single character recognition method where focus lies on ensuring the additive property such that it can easily be extended to be used for on-line cursive character recogni- tion in a well-defined way. Various experiments with the new methodology have been conducted on the UNIPEN database with promising results.