SHAPE VECTORS - AN EFFICIENT PARAMETRIC REPRESENTATION FOR THE SYNTHESIS AND RECOGNITION OF HAND SCRIPT CHARACTERS

Earlier work by the author has established: (i) that cursive script can be synthesised out of individual characters by using polynomial merging functions which satisfy boundary conditions of continuity of the displacement functions x(t) and y(t) for each character and their first and second derivatives; and (ii) that the procedure lends itself to a Bezier curve based formulation. This was done since cursive writing avoids discontinuities (of shape) between individual characters as well as discontinuities in pen movement. We show here that even individual characters can be synthesised out of more primitive elements by using the same merging functions. We choose directed straight lines which we call shape vectors as basic elements for this. Script characters generally have shapes which are essentially straight segments alternating with 'bends' or regions of relatively high curvature. For a character with n bends, we need only n + 1 shape vectors. Thus, each script character needs only three to seven shape vectors, depending on its complexity. The ''character generation'' shape vectors are derived from the original character by means of a simple procedure that identifies comparatively straight regions in it. These are then approximated to straight lines by linear regression and positioned to be tangential to the original curve. The synthesised version of this character is obtained by 'merging' or concatenating these vectors. The close fit between the original and resynthesised characters demonstrates that the shape vectors adequately characterise their identities and shapes. Data reduction ratios in the range of 15 to 25 are thus possible. This method thus shows good promise as a possible basis for script character recognition, and a recognition scheme based on it has yielded an accuracy of 94% for a vocabulary size of 67 words.