Using sequences of local minima/maxima for on-line cursive letter recognition

A recognition process is based on a comparison of an object to be identified (recognized) with objects known by the subject performing the recognition. In case of machine letter recognition this process is performed by a computer and software implemented on it and the objects to be recognized are letters. For on-line recognition letters are supplied as a set of consecutive coordinates of the pen tip position as it moves along the path forming the shape of a letter. The recognition algorithm may attempt to compare the input shape directly with ones “known” to it, or it may attempt to extract certain features which characterize the shape and compare only those with the reference feature sets “known”. The latter approach has the potential of being more flexible and robust, as the well chosen feature set will tolerate various distortions of shapes/patterns to be recognized.