Adaptive soft k-nearest-neighbour classifiers

A novel classifier is introduced to overcome the limitations of the k-NN classification systems. It estimates the posterior class probabilities using a local Parzen window estimation with the k-nearest-neighbour prototypes (in the Euclidean sense) to the pattern to classify. A learning algorithm is also presented to reduce the number of data points to store. Experimental results in two hand-written classification problems demonstrate the potential of the proposed classification system.     

Display text segmentation after learning best-fitted OCR binarization parameters

In this paper text segmentation in generic displays is proposed through learning the best binarization values for a commercial optical character recognition (OCR) system. The commercial OCR is briefly introduced as well as the parameters that affect the binarization for improving the classification scores. The purpose of this work is to provide the capability to automatically evaluate standard textual display information, so that tasks that involve visual user verification can be performed without human intervention. The problem to be solved is to recognize text characters that appear on the display, as well as the color of the characters’ foreground and background. The paper introduces how the thresholds are learnt through: (a) selecting lightness or hue component of a color input cell, (b) enhancing the bitmaps’ quality, and (c) calculating the segmentation threshold range for this cell. Then, starting from the threshold ranges learnt at each display cell, the best threshold for each cell is gotten. The input and output data sets for testing the algorithms proposed are described, as well as the analysis of the results obtained.     

手写混合字符集识别的多特征多级分类器设计

针对常用的银行汉字和阿拉伯数字混合字符集的识别,提出了依据不同的分类要求，分别选取不同的分类特征，并采用先聚类再用多层感知器(MLP)神经网络分类的多级分类器进行识别的设计方法。实验结果表明，该方法用于手写体混合字符集的识别是行之有效的。     

Comparison of kernel estimators,perceptrons and radial-basis functions for OCR and speech classification

We compare kernel estimators, single and multi-layered perceptrons and radial-basis functions for the problems of classification of handwritten digits and speech phonemes. By taking two different applications and employing many techniques, we report here a two-dimensional study whereby a domain-independent assessment of these learning methods can be possible. We consider a feed-forward network with one hidden layer. As examples of the local methods, we use kernel estimators like k-nearest neighbour (k-nn), Parzen windows, generalised k-nn, and Grow and Learn (Condensed Nearest Neighbour). We have also considered fuzzy k-nn due to its similarity. As distributed networks, we use linear perceptron, pairwise separating linear perceptron and multi-layer perceptrons with sigmoidal hidden units. We also tested the radial-basis function network, which is a combination of local and distributed networks. Four criteria are taken for comparison: correct classification of the test set; network size; learning time; and the operational complexity. We found that perceptrons, when the architecture is suitable, generalise better than local, memory-based kernel estimators, but require a longer training and more precise computation. Local networks are simple, leant very quickly and acceptably, but use more memory.     

面向中文矢量笔迹中单字详细分类的特征和方法比较*

在识别矢量笔迹文本时,不同类型单字需要采用不同识别器,确定详细类别是单字识别的前提。对实际中文矢量笔迹文本中单字进行汉字、标点、数字、字母和单词的详细分类,提出了自身和相对（包括近邻和同行）特征,选用决策树、逻辑模型树、贝叶斯网络和支持向量机四种分类器。针对大量实际数据,测试和比较了多种特征和分类器的性能。实验表明,近邻单字的组合特征具有较好的分类能力,支持向量机对各种单字均有较好分类性能。     

基于距离加权的自适应字线分离算法

提出一种基于距离加权的自适应字线分离算法。应用一定的启发式规则，计算表格线上像素点的权值，将权值与阈值相比较来判断该点是否为字符上的点，其中权值和阈值根据具体表格自动确定。该算法与表格线检测方法无关，且易于实现。实验结果表明，可以很好地处理字线交叠问题，提高了表格识别的正确率。     

基于MQDF的英文OCR多模板分类器

针对进一步提高英文OCR分类器的鲁棒性进行了研究,结合传统的单模板MQDF分类器和多模板欧氏距离分类器各自的优点,提出了一种新的基于MQDF的多模板分类器设计方法。与传统分类器的对比测试证明,该文提出的这种新的分类器能够有效地提高多体英文OCR字符的单字首选正确率,并对低质量文本中的模糊和断裂字符也能保持很高的识别率。     

The impact of new hardware on OCR designs

This paper describes the opportunities available to radically improve the performance cost ratio of OCR machines by taking advantage of new hardware which is either here, or can be made available by inter industry cooperation. Major discussions center around:(1) The page width self-scanned-array (1872 photodiodes) which can be used to greatly reduce the number of mechanical operations needed to scan a multi-line document: (2) The much reduced cost of semiconductor memory (particularly the charge coupled devices) which allows the use of memory to eliminate double scanning as well as other awkward and embarrassing operations: (3) A proposal for an LSI correlator which can provide substantial gains in all performance and cost categories, as well as providing prospects for the modest application of developments in learning theory. The leadership to provide such a correlator may have to spring from the OCR industry, including agreement on certain standards.     

一种组合特征抽取的新方法

该文提出了一种基于特征级融合的特征抽取新方法，首先，给出了一种合理的特征融合策略，即利用复向量给出组合特征的表示，将特征空间从实向量空间拓广到复向量空间，然后，发展了具有统计不相关性的鉴别分析的理论，并将其用于复向量空间内最优鉴别特征的抽取，最后，在Concordia大学的CENPARMI手写体阿拉伯数字数据库以及南京理工大学NUST603HW手写汉字库上的试验结果表明，所提出的组合特征抽取方法不仅具有很强的维数压缩能力，而且较大幅度地提高了识别率。     

Reduction of processing time for optimal and quadratic discriminant analyses

A fast algorithm is presented for optimal discriminant analysis and quadratic discriminant analysis. In this algorithm, the discriminant function of an input feature vector for each category is calculated via a monotonically increasing sequence, and when the sequence value exceeds a certain value, then you can assert that the current category cannot be the classification result and omit the redundant calculation of the remaining terms for the category, thus making the calculation faster. Applying this algorithm to the recognition experiment on handwritten characters, we could reduce the processing time to 4% of the conventional simple method. Since both discriminant analyses assume the normal distribution of the features, disnormality contained in real-world data affects the accuracy of the two discriminant analyses. We also compared the accuracy performances of the two discriminant analyses using real-world data and artificial data.     

Adaptive fuzzy particle swarm optimization for global optimization of multimodal functions

This paper proposes an adaptive fuzzy PSO (AFPSO) algorithm, based on the standard particle swarm optimization (SPSO) algorithm. The proposed AFPSO utilizes fuzzy set theory to adjust PSO acceleration coefficients adaptively, and is thereby able to improve the accuracy and efficiency of searches. Incorporating this algorithm with quadratic interpolation and crossover operator further enhances the global searching capability to form a new variant, called AFPSO-QI. We compared the proposed AFPSO and its variant AFPSO-QI with SPSO, quadratic interpolation PSO (QIPSO), unified PSO (UPSO), fully informed particle swarm (FIPS), dynamic multi-swarm PSO (DMSPSO), and comprehensive learning PSO (CLPSO) across sixteen benchmark functions. The proposed algorithms performed well when applied to minimization problems for most of the multimodal functions considered.     

一种自适应加权变形的2DPCA人脸识别方法

为了充分利用图像矩阵的局部信息和更多的鉴别信息,以提高2DPCA的识别率,提出了一种自适应加权变形的2DPCA人脸识别方法.该方法将人脸图像矩阵分块,然后利用变形的2DPCA方法提取特征,接着自适应地计算每个分块在分类中的权值,最后根据类别的权值大小进行分类.在ORL人脸库中进行的实验研究表明,该方法在正确识别率和识别...     

Regularized discriminant analysis for the small sample size problem in face recognition

It is well-known that the applicability of both linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) to high-dimensional pattern classification tasks such as face recognition (FR) often suffers from the so-called “small sample size” (SSS) problem arising from the small number of available training samples compared to the dimensionality of the sample space. In this paper, we propose a new QDA like method that effectively addresses the SSS problem using a regularization technique. Extensive experimentation performed on the FERET database indicates that the proposed methodology outperforms traditional methods such as Eigenfaces, direct QDA and direct LDA in a number of SSS setting scenarios.     

支票大写金额的印刷体与手写体辨识

针对银行支票大写金额识别工作中的手写体和印刷体辨识，抽取了4组分类特征，较好地体现了印刷体字符和手写体字符的区别。并且在构造分类器时，介绍了求取各分类特征对识别结果影响大小的权值向量的方法。在银行提供的大量真实票据上进行了验证，并且应用于实际的银行自动支票处理系统，取得了很好的效果。     

An intelligent character recognizer for Telugu scripts using multiresolution analysis and associative memory

The present work is an attempt to develop a robust character recognizer for Telugu texts. We aim at designing a recognizer, which exploits the inherent characteristics of the Telugu Script. Our proposed method uses wavelet multi-resolution analysis for the purpose extracting features and associative memory model to accomplish the recognition tasks. Our system learns the style and font from the document itself and then it recognizes the remaining characters in the document. The major contribution of the present study can be outlined as follows. It is a robust OCR system for Telugu printed text. It avoids feature extraction process and it exploits the inherent characteristics of the Telugu character by a clever selection of Wavelet Basis function, which extracts the invariant features of the characters. It has a Hopfield-based Dynamic Neural Network for the purpose of learning and recognition. This is important because it overcomes the inherent difficulties of memory limitation and spurious states in the Hopfield Network. The DNN has been demonstrated to be efficient for associative memory recall. However, though it is normally not suitable for image processing application, the multi-resolution analysis reduces the sizes of the images to make the DNN applicable to the present domain. Our experimental results show extremely promising results.     

Adaptive model for decision making

Learning of an adaptive model in terms of the creation of n-dimensional bodies is presented. N-dimensional bodies are used to partition the space Rⁿ of measurable pattern parameters in order to separate patterns of one answer, e.g. class or type of disease, from those of other ones.

The model, when applied medically in the prognosis of acute pancreatitis and in the diagnosis of disseminated cancer of unknown origin, showed good results.     

基于归一化背景方向特征的脱机手写汉字识别

对4方向背景方向特征进行了改进,提出了8方向背景特征描述方法。与4方向背景方向特征描述方法相比,改进后的特征描述方法可以从0°、45°、90°、135°、180°、225°、270°、315°共8个方向来对汉字图像进行考察,从而进一步提高描述的精度。此外,为了消除笔划粗细的影响,还对背景方向特征进行了归一化处理。实验结果表明改进后的归一化8方向背景方向特征具有更高的识别精度。     

A New Algorithm for Generalized Optimal Discriminant Vectors

A study has been conducted on the algorithm of solving generalized optimal set of discriminant vectors in this paper.This paper proposes an analytical algorithm of solving generalized optimal set of discriminant vectors theoretically for the first time.A lot of computation time can be saved because all the generalized optimal ests of discriminant vectors can be obtained simultaneously with the proposed algorithm,while it needs no iterative operations .The proposed algorithm can yield a much higher recognition rate.Furthermore,the proposed algorithm overcomes the shortcomings of conventional human face recognition algorithms which were effective for small sample size problems only.These statements are supported by the numerical simulation experiments on facial database of ORL.     

完备自适应近邻图嵌入的局部鉴别投影算法

针对基于自适应近邻图嵌入的局部鉴别投影算法（Neighborhood graph embedding based local adaptive discriminant analysis, LADP ）仅仅利用局部类内离差矩阵主元空间的鉴别信息而丢失了其零空间内大量鉴别信息的不足,结合全空间的基本思想提出了完备的基于自适应近邻图嵌入的局部鉴别投影算法（ Complete LADP,CLADP）。在局部类内离差矩阵的零空间内,通过最大化局部类间离差矩阵提取不规则鉴别特征,在局部类间离差矩阵的主元空间内,通过最大化局部类间离差矩阵的同时最小化局部类 内离差矩阵提取规则鉴别特征,最后将不规则鉴别特征和规则鉴别特征串联形成CLADP特征。在ORL,Yale以及PIE人脸库上的人脸识别实验结果证明了CLADP的有效性。     

手写字符识别的非线性主动判别函数

提出了用于手写字符识别的非线性主动判别函数,是线性主动判别函数在手写字符非线性变化情况下的推广。该方法利用Kernel PCA分析捕捉和表示这种非线性变化。将输入空间非线性映射为特征空间,在特征空间的主子空间中生成最优主动原型模板,其与字符特征向量在特征空间主子空间的投影之间的距离即为非线性主动判别函数;同时,基于最小分类错误准则对该函数进行了优化。实验结果表明,非线性主动判别函数获得了比线性主动判别函数更高的识别率。