AUTOMATIC DETECTION OF HANDWRITING FORGERY USING A FRACTAL NUMBER ESTIMATE OF WRINKLINESS

Abstract

We investigate the detection of handwriting forged by novices. To facilitate document examination it is important to develop an automated system to identify forgeries, or at least to identify those handwritings that are likely to be forged. Because forgers often carefully copy or trace genuine handwriting, we hypothesize that good forgeries — those that retain the shape and size of genuine writing — are usually written more slowly and are therefore wrinklier (less smooth) than genuine writing. From online handwriting samples we find that the writing speed of the good forgeries is significantly slower than that of the genuine writings. From corresponding offline samples we find that the wrinkliness of the good forgeries is significantly greater than that of the genuine writings, showing that this feature can help identify candidate forgeries from scanned documents. Using a total of eight handwriting distance features, including the wrinkliness feature, we train a neural network to achieved 89% accuracy on detecting forged handwriting on test samples from ten writers.