Writer Identification for Offline Handwritten Kanji without using Character Recognition Features
Ayumu Soma, Masayuki Arai · 2013
Most research on writer identification in the case of offline handwritten Kanji characters uses character recognition features.In this paper, we assume the following is representative of the writers' style: the start point, the end point, the angle of each stroke composing a Kanji character, and the size and position of the Kanji character.The experimental results show that the identification rates are 95.2% without rejection and 99.6% with 10% rejection for four Kanji characters written by one hundred writers.