Parameter calibration for synthesizing realistic-looking variability in offline handwriting

Being motivated by the widely accepted principle that the more training data we have, the better performance the recognition system has, we conducted experiments asking human subjects to do test on a mixture of real English handwritten textlines and textlines altered from existing handwriting with various distortion degrees. The idea of generating synthetic handwriting is based on a perturbation method by T. Varga and H. Bunke that distorts an entire textline. There are two purposes of our experiments. First, we want to calibrate optimal distortion parameter settings for Varga and Bunke’s perturbation model. Second, we intend to compare the effects of parameter settings on different writing styles, block, cursive and mixed. From the preliminary experimental results, we determined appropriate ranges for parameter amplitude, and found that parameter settings should change for different handwriting styles. Once the proper parameter settings are found, we will generate large amount of training and testing sets for building better off-line handwriting recognition systems.