Research And Implementation Of An On-line Handwriting Recognition System Based On Neural Network

Computer recognition of Chinese characters is an important branch of pattern recognition. On-Line Chinese Character Recognition (OLCCR) system accords with the nature habit when people write, so it can be easily accepted by people. Now researchers have done much research work on it and proposed many pattern recognition algorithms. On-line handwriting recognition has been basically realized, however most studies are carried out for one kind of handwritten patterns. Although these systems can achieve higher recognition rate, the users need to switch the input mode before writing. The performance should be further improved.With the capabilities of parallel computing, fault tolerance, self-learning and classification, Artificial Neural Network is paid full attention and used in the field of pattern recognition. DHNN is used in this paper. The learning samples are firstly preprocessed including binarization, normalization and interpolation. Secondly their pixel features are extracted. Then DHNN is established. Finally the handwritten test samples are input into the network. It can be seen from the simulation results that the system can recognize the incomplete characters and characters with noises. The simulation results show that the recognition results using Hopfield are better than using BP.