Prototype Integration in Off-line Handwriting Recognition Adaptation

Writer adaptation or specialization is the adjustment of handwriting recognition algorithms to a specific writer’s style of handwriting. Such adjustment yields significantly improved recognition rates over counterpart general recognition algorithms. We present a discussion of a method of prototype integration for writer adaptation and evaluate the results on English and Arabic datasets. The writer adaptation model we use is an iterative bootstrapping model which adapts a writer-independent model to a writer-dependent model using a small number of words achieving a large recognition rate increase in the process, utilizing a confidence weighting method which generates better results by weighting words based on their length. The Arabic testing set consisting of about 100 pages of handwritten text and was able to improve its recognition rate by over 10% using adaptation. The English testing set consisted of a subset of 3000 pages of handwritten letters written by 1000 authors. Similar to the Arabic set, a corresponding increase in performance was observed.