基于神经网络信息融合的印刷体字符识别研究

Research of printed character recognition based on neural network information fusion

针对印刷体字符识别,提出一种基于神经网络信息融合的方法。在对待识别目标提取特征后,分别采用2种反向传播算法的改进算法和遗传算法构造神经网络分类器模型,并进行网络的训练和识别工作。通过实验数据着重分析和比较了3种算法的特点,将此3种分类器得出的分类结果进行决策级的信息融合,最终得出识别结果。实验结果表明,此方法简单可行,具有较高的鲁棒性和识别率。

Considering the status of the printed character recognition, this paper proposed a new approach which based on the neural network information fusion. After extracting the features of the target, using two back-propagation algorithm and genetic algorithms to build the neural network models to train and recognize the targets. Analyzes and compares the characteristic of the three algorithms through experiment data, take identification results from the three classifiers to the decision-making level information fusion and get the final results. The experiments results show that this method is simple and feasible with high robustness and recognition rate.