阅读认知模式下的图案识别方法研究

本文模拟人类对图案的认知识别机理，提出了一种基于阅读认知模式的特征提取方法，提取基于视觉信息的图案特征，并提出了一种基于基元拓扑关系建模的通用图案识别方法。利用滑动窗来实现对人类认知图案机制的模拟，通过滑动窗的滑动过程完成对图案局部结构特征提取以及空间拓扑关系的构建。在图案识别建模方法中，采用了人工神经网络和隐马尔科夫模型相结合的混合识别模型，利用人工神经网络的强大计算能力完成基元建模，结合隐马尔科夫模型的强大的处理时序数据的优势，实现了图案的整体拓扑结构建模。实验结果验证了本文提出的图案识别方发的有效性和通用性。 

Research of image recognition method based on reading cognitive model

This paper simulates the human being's cognitive mechanism of pattern recognition. Based on the reading cognition model, a feature extraction method is proposed to calculate the image features from the visual information point of view. A universal pattern recognition framework is constructed through the modeling of the topological relationship of primitives. A sliding window approach is applied to simulate the human pattern cognition mechanism. The sliding process is used to extract the local structure features and assemble the topological relationship at the meantime. In this paper, the recognition model is a hybrid of the predictive artificial neural network ANN and the hidden markov model HMM. ANNs are used to model the primitives of the patterns depending on their supper computing ability, and the HMM is used to model the pattern's overall topological structure according to its strong ability of time series data processing. The experimental results verified the effectiveness and versatility of the proposed method.