Using neural nets to recognize handwritten/printed characters
C.C. Chiang, H.C. Fu · 2002
A handwritten character recognition system implemented by a stochastic neural net (SNN) is presented. The learning process in this neural model incorporates the stochastic information extracted from the trained characters. The merit of SNN lies in the fast learning speed obtained through an online learning algorithm, in contrast to offline learning algorithms. This SNN has been applied to design an experimental adaptive character recognition system. This system can recognize handwritten/printed English and Chinese characters. According to preliminary experimental results, the recognition rates are about 90 approximately 94% and 80 approximately 85% for English and Chinese characters, respectively. The system is capable of learning while recognizing new patterns.>