基于仿反馈机制的脱机手写体汉字认知模型

An Off-line Handwritten Chinese Character Cognitive Model Based on Simulated Feedback Mechanism

针对已有认知模型面对不同样本采用相同认知需求和固定双树复小波变换(Dual-Tree Complex Wavelet Transform,DTCWT)特征空间的缺陷,仿人面对不同认知需求时自适应调节特征空间反复推敲比对的认知特点,探索一种具有仿反馈调节机制的脱机手写体汉字智能认知模型。首先,提出了一种具有仿反馈调节机制的智能认知模型;第二,分析脱机手写体汉字样本的认知需求,构建不同认知需求下的优化DTCWT特征子空间与分类认知准则;第三,定义认知结果评价准则,通过分析未认知样本的认知需求,自适应调节新认知需求下优化特征子空间和分类认知准则,集成多认知需求下的优化特征子空间构建优化压缩DTCWT特征空间。对GB2312-80手写体汉字样本库进行仿真实验,实验结果表明该方法的优越性。

For the drawbacks of existing cognitive models with same cognitive demanding and constant feature space constructed by DTCWT for various samples, an intelligent off-line handwritten Chinese character cognitive model with simulated feedback adjustment mechanism is explored in this paper, to simulate the human cognitive process of the adaptive adjusting feature space to repeat intercomparison and deliberately refine with various cognitive demanding. Firstly, an intelligent cognitive model with simulated feedback adjustment mechanism is proposed. Secondly, the cognitive demanding of samples is analyzed to establish the optimized DTCWT feature subspace and classified cognitive rules for various sample cognitive demanding.Thirdly, the evaluation criteria of cognitive results is defined to adaptively adjust the optimized feature subspace and classified cognitive rules based on the new cognitive demanding from the falsely cognitive samples. The optimized compact DTCWT feature space is established by integrating various optimized feature subspaces of multi-cognitive demanding. The experimental results based on GB2312-80 handwritten Chinese sample library show the superiority of our method.