Arabic Word Recognition by Classifiers and Context

Given the number and variety of methods used for handwriting recognition, it has been shown that there is no single method that can be called the "best". In recent years, the combination of different classifiers and the use of contextual information have become major areas of interest in improving recognition results. This paper addresses a case study on the combination of multiple classifiers and the integration of syntactic level information for the recognition of handwritten Arabic literal amounts. To the best of our knowledge, this is the first time either of these methods has been applied to Arabic word recognition. Using three individual classifiers with high level global features, we performed word recognition experiments. A parallel combination method was tested for all possible configuration cases of the three chosen classifiers. A syntactic analyzer makes a final decision on the candidate words generated by the best configuration scheme. The effectiveness of contextual knowledge integration in our application is confirmed by the obtained results.     

Improving the classification accuracy of streaming data using SAX similarity features

The classification accuracy of time series is highly dependent on the quality of used features. In this study, features of new type, called SAX (Symbolic Aggregate approXimation) similarity features, are presented. SAX similarity features are a combination of the traditional statistical number-based and the template-based classification. SAX similarity features are obtained from the data of the time window by first transforming the time series into a discrete presentation using SAX. Then the similarity between this SAX presentation and predefined SAX templates are calculated, and these similarity values are considered as SAX similarity features. The functioning of these features was tested using five different activity data sets collected using wearable inertial sensors and five different classifiers. The results show that the recognition rates calculated using SAX similarity features together with traditional features are much better than those obtained employing traditional features only. In 20 tested cases out of 23, the improvement is statistically significant according to the paired t-test.     

一种手写图文分离方法

手写识别作为改善人机交互的技术之一已经变得越来越重要,涌现了大量对手写文字和手绘图形的研究工作,而作为手写识别的一个重要部分,对图形和文本的分类工作一直没有获得足够的重视。本文基于开源数据挖掘工具Weka设计并实现一种手写图文分离方法,基于LogitBoost、Random Forest和LADTree三种不同分类器的测试结果表明,Logit-Boost的综合分类效果最好。通过联合3个分类器能够实现精确的图形判定,但文本的分类效率则受限于分类效果最差的分类器。同时基于信息增益评估结果,分析了不同特征对图文分类的影响。     

Application of multi-layer perceptron neural networks to vision problems

This paper discusses the application of a class of feed-forward Artificial Neural Networks (ANNs) known as Multi-Layer Perceptrons(MLPs) to two vision problems: recognition and pose estimation of 3D objects from a single 2D perspective view; and handwritten digit recognition. In both cases, a multi-MLP classification scheme is developed that combines the decisions of several classifiers. These classifiers operate on the same feature set for the 3D recognition problem whereas different feature types are used for the handwritten digit recognition. The backpropagationlearning rule is used to train the MLPs. Application of the MLP architecture to other vision problems is also briefly discussed.     

Boundary control,stabilization and zero-pole dynamics for a non-linear distributed parameter system

In this work we show that the now standard lumped non-linear enhancement of root-locus design still persists for a non-linear distributed parameter boundary control system governed by a scalar viscous Burgers' equation. Namely, we construct a proportional error boundary feedback control law and show that closed-loop trajectories tend to trajectories of the open-loop zero dynamics as the gain parameters are increased to infinity. We also prove a robust version of this result, valid for perturbations by an unknown disturbance with arbitrary L² norm. For the controlled Burgers' equation forced by a disturbance we prove that, for all initial data in L²(0, 1), the closed-loop trajectories converge in L²(0, 1), uniformly in t∈[0, T] and in H¹(0, 1), uniformly in t∈[t₀, T] for any t₀>0, to the trajectories of the corresponding perturbed zero dynamics. We have also extended these results to include the case when additional boundary controls are included in the closed-loop system. This provides a proof of convergence of trajectories in case the zero dynamics is replaced by a non-homogeneous Dirichlet boundary controlled Burgers' equation. As an application of our convergence of trajectories results, we demonstrate that our boundary feedback control scheme provides a semiglobal exponential stabilizing feedback law in L², H¹ and L^∞ for the open-loop system consisting of Burgers' equation with Neumann boundary conditions and zero forcing term. We also show that this result is robust in the sense that if the open-loop system is perturbed by a sufficiently small non-zero disturbance then the resulting closed-loop system is ‘practically semiglobally stabilizable’ in L²-norm. Copyright © 1999 John Wiley & Sons, Ltd.     

A geometric property of control systems with states in a Banach space

In this paper we are concerned with admissible trajectories x(·) of infinite-dimensional linear systems, defined on an interval [0, t₁] and such that x(0)−x(t₁) is small. We show that these trajectories satisfy certain geometric properties.     

Directional features in online handwriting recognition

The selection of valuable features is crucial in pattern recognition. In this paper we deal with the issue that part of features originate from directional instead of common linear data. Both for directional and linear data a theory for a statistical modeling exists. However, none of these theories gives an integrated solution to problems, where linear and directional variables are to be combined in a single, multivariate probability density function. We describe a general approach for a unified statistical modeling, given the constraint that variances of the circular variables are small. The method is practically evaluated in the context of our online handwriting recognition system frog on hand and the so-called tangent slope angle feature. Recognition results are compared with two alternative modeling approaches. The proposed solution gives significant improvements in recognition accuracy, computational speed and memory requirements.     

基于多分类器组合的笔迹验证

运用多分类器组合技术和模糊技术将多种笔迹鉴别方法按一定规则进行融合，针对笔迹鉴别中的笔迹验证问题进行应用。实验结果表明，融合后笔迹验证准确率有大幅的提高     

A paradigm for handwriting-based intelligent tutors

This paper presents the interaction design of, and demonstration of technical feasibility for, intelligent tutoring systems that can accept handwriting input from students. Handwriting and pen input offer several affordances for students that traditional typing-based interactions do not. To illustrate these affordances, we present evidence, from tutoring mathematics, that the ability to enter problem solutions via pen input enables students to record algebraic equations more quickly, more smoothly (fewer errors), and with increased transfer to non-computer-based tasks. Furthermore our evidence shows that students tend to like pen input for these types of problems more than typing. However, a clear downside to introducing handwriting input into intelligent tutors is that the recognition of such input is not reliable. In our work, we have found that handwriting input is more likely to be useful and reliable when context is considered, for example, the context of the problem being solved. We present an intelligent tutoring system for algebra equation solving via pen-based input that is able to use context to decrease recognition errors by 18% and to reduce recognition error recovery interactions to occur on one out of every four problems. We applied user-centered design principles to reduce the negative impact of recognition errors in the following ways: (1) though students handwrite their problem-solving process, they type their final answer to reduce ambiguity for tutoring purposes, and (2) in the small number of cases in which the system must involve the student in recognition error recovery, the interaction focuses on identifying the student’s problem-solving error to keep the emphasis on tutoring. Many potential recognition errors can thus be ignored and distracting interactions are avoided. This work can inform the design of future systems for students using pen and sketch input for math or other topics by motivating the use of context and pragmatics to decrease the impact of recognition errors and put user focus on the task at hand.     

Tool-path generation of planar NURBS curves

It has been widely used in CAD field for many years and gradually applied in CAM area with the prevalence of NURBS interpolator equipped in CNC controllers. But few of them provide the tool radius compensation function. In order to achieve the goal of generating tool-path, an algorithm was presented to offset NURBS curves by an optimum process for CAD/CAM systems in this paper. NURBS format is ideal for HSM applications, but not all NURBS outputs are equal and standard. Basically, there are two different ways to generate NURBS tool-paths; one is to fit a NURBS curve to the conventional tool-path output, the other one is to generate a NURBS tool-path from the start. The main targets for the tool-path of this paper are: (1) To keep a constant distance d between progenitor curve C(t) and offset curve C_d(t) on the normal direction of C(t); (2) to alternate the order k of the basis function in offset curve C_d(t); (3) to oscillate the number of control points of offset curve C_d(t) and compare it with progenitor curve C(t). In order to meet the tolerance requirements as specified by the design, this study offsets the NURBS curves by a pre-described distance d. The principle procedure consists of the following steps: (1) construct an evaluating bound error function; (2) sample offset point-sequenced curves based on first derivatives; (3) give the order of NURBS curve and number of control points to compute all initial conditions and (4) optimize the control points by a path searching algorithm.     

Efficient implementation of class-based decomposition schemes for Naïve Bayes

Previous studies have shown that the classification accuracy of a Naïve Bayes classifier in the domain of text-classification can often be improved using binary decompositions such as error-correcting output codes (ECOC). The key contribution of this short note is the realization that ECOC and, in fact, all class-based decomposition schemes, can be efficiently implemented in a Naïve Bayes classifier, so that—because of the additive nature of the classifier—all binary classifiers can be trained in a single pass through the data. In contrast to the straight-forward implementation, which has a complexity of O(n?t?g), the proposed approach improves the complexity to O((n+t)?g). Large-scale learning of ensemble approaches with Naïve Bayes can benefit from this approach, as the experimental results shown in this paper demonstrate.     

联机手写英文识别

英文手写识别既是模式识别问题又是一个搜索问题,其计算复杂度非常高．针对这个问题,提出了一些新的方法来解决联机英文手写识别问题．首先,运用一些先验知识对所获得的基线进行调整;其次,针对英文识别过程的复杂度太高,提出了快速有效的解码算法——主笔划分层构筑Viterbi算法;最后,将多阶段的分类信息融合到决策过程中,较大幅度地提高了系统的性能．在Unipen训练集和实验室数据集上测试取得了良好的效果．     

基于支持向量机的思维脑电信号特征分类研究

探索一种实用的基于想象运动思维脑电的脑-机接口(BCI)方式,为实现BCI应用奠定比较坚实的理论和实验基础。对6名受试者进行三种不同时段(箭头出现2s、1s和0s后提示按键)情况下想象左右手运动思维作业的信号采集实验,利用小波变换和支持向量机对实验数据进行离线处理。对三种情况下的延缓时间△t0、△t1和△t2分析发现:△t0与△t1和△t2之间都有显著性差别(p<0.05),而△t1与△t2之间没有显著差别(p>0.05);平均分类正确率分别达到68.00%、80.00%和56.67%(p<0.05);实际按键前0.5~1s左右,想象左右手运动的思维脑电特征信号都发生了明显改变。通过合理的实验设计获取的信号有助于识别正确率的提高,为BCI系统中思维任务的特征提取与识别分类提供了新思路和方法。     

On optimal pairwise linear classifiers for normal distributions: the d-dimensional case

We consider the well-studied pattern recognition problem of designing linear classifiers. When dealing with normally distributed classes, it is well known that the optimal Bayes classifier is linear only when the covariance matrices are equal. This was the only known condition for classifier linearity. In a previous work, we presented the theoretical framework for optimal pairwise linear classifiers for two-dimensional normally distributed random vectors. We derived the necessary and sufficient conditions that the distributions have to satisfy so as to yield the optimal linear classifier as a pair of straight lines.In this paper we extend the previous work to d-dimensional normally distributed random vectors. We provide the necessary and sufficient conditions needed so that the optimal Bayes classifier is a pair of hyperplanes. Various scenarios have been considered including one which resolves the multi-dimensional Minsky’s paradox for the perceptron. We have also provided some three-dimensional examples for all the cases, and tested the classification accuracy of the corresponding pairwise-linear classifier. In all the cases, these linear classifiers achieve very good performance. To demonstrate that the current pairwise-linear philosophy yields superior discriminants on real-life data, we have shown how linear classifiers determined using a maximum-likelihood estimate (MLE) applicable for this approach, yield better accuracy than the discriminants obtained by the traditional Fisher's classifier on a real-life data set. The multi-dimensional generalization of the MLE for these classifiers is currently being investigated.     

LoGID: An adaptive framework combining local and global incremental learning for dynamic selection of ensembles of HMMs

In this work, we propose the LoGID (Local and Global Incremental Learning for Dynamic Selection) framework, the main goal of which is to adapt hidden Markov model-based pattern recognition systems during both the generalization and learning phases. Given that the baseline system is composed of a pool of base classifiers, adaptation during generalization is performed through the dynamic selection of the members of this pool that best recognize each test sample. This is achieved by the proposed K-nearest output profiles algorithm, while adaptation during learning consists of gradually updating the knowledge embedded in the base classifiers, by processing previously unobserved data. This phase employs two types of incremental learning: local and global. Local incremental learning involves updating the pool of base classifiers by adding new members to this set. The new members are created with the Learn++ algorithm. Global incremental learning, in contrast, consists of updating the set of output profiles used during generalization. The proposed framework has been evaluated on a diversified set of databases. The results indicate that LoGID is promising. For most databases, the recognition rates achieved by the proposed method are higher than those achieved by other state-of-the-art approaches, such as batch learning. Furthermore, the simulated incremental learning setting demonstrates that LoGID can effectively improve the performance of systems created with small training sets as more data are observed over time.     

Combining diverse systems for handwritten text line recognition

In this paper, we present a recognition system for on-line handwritten texts acquired from a whiteboard. The system is based on the combination of several individual classifiers of diverse nature. Recognizers based on different architectures (hidden Markov models and bidirectional long short-term memory networks) and on different sets of features (extracted from on-line and off-line data) are used in the combination. In order to increase the diversity of the underlying classifiers and fully exploit the current state-of-the-art in cursive handwriting recognition, commercial recognition systems have been included in the combined system, leading to a final word level accuracy of 86.16%. This value is significantly higher than the performance of the best individual classifier (81.26%).     

Characterization of analytic phase signals

In many cases, a real-valued signal χ(t) may be associated with a complex-valued signal a(t)e^iθ(t), the analytic signal associated with χ(t) with the characteristic properties χ(t) = a(t) cosθ(t) and H(a(·)cosθ(·))(t) = a(t)sinθ(t). Using such obtained amplitude-frequency modulation the instantaneous frequency of χ(t) at the time t₀ may be defined to be θ′(t₀), provided θ′(t₀) ≥ 0. The purpose of this note is to characterize, in terms of analytic functions, the unimodular functions F(t) = C(t) + iS(t),C²(t) + S² (t) = 1, a.e., that satisfy HC(t) = S(t). This corresponds to the case a(t) ≡ 1 in the above formulation. We show that a unimodular function satisfies the required condition if and only if it is the boundary value of a so called inner function in the upper-half complex plane. We also give, through an explicit formula, a large class of functions of which the parametrization C(t) = cosθ(t) is available and the extra condition θ′(t) ≥ 0, a.e. is enjoyed. This class of functions contains Blaschke products in the upper-half complex plane as a proper subclass studied by Picinbono in [1].     

多分类器组合研究

文章提出了一种多分类器的组合方法,它利用了参与组合的分类器提供的度量层次上的两类信息:对训练样本的决策信息;对待识样本的决策信息。首先对这两类信息进行集成,进而给出了组合分类器的判定规则。用该方法对手写体汉字作分类识别,实验结果显示,较之其它几种方法,它有更高的正确识别率。