用于手写字符识别的SA-ART神经网络模型

针对前馈 BP网络和 ART网络对手写字符识别的不足以及人的认知机制 ,在大量实践和理论分析的基础上 ,提出具有选择注意机制的 ART模型 SA- ART(Selective Attention-Adaptive Resonance Theory) .经实际检测 ,识别精度有明显提高 .

A Kind of SA-ART Neural Network Mode Used in the Identification of Handwriting Characters

The ART-network and the BP-network are usually used in identifying handwriting characters. Although the ART-network in identifying handwriting characters is more powerful than the BP-network, the improvement is limited and the accuracy of the identification is much lower than that excepted. On the basis of a lot of practical and theoretical experiments, we brought forward one kind of ART modes called SA-ART (Selective Attention-Adaptive Resonance Theory) aiming at the BP-network's and the ART-network's deficiencies and the human cognition mechanism. The SA-ART has the selective ability. After making plenty of practical tests, it was found that the accuracy of the AR-SART is much higher than before.