Search

This paper introduces a new topic and research of geometric classifier ensemble learning using two types of objects: classifier prediction pairwise matrix (CPPM) and decision profiles (DPs). Learning from CPPM requires using Riemannian manifolds (R-manifolds) of symmetric positive definite (SPD) matrices. DPs can be used to build a Grassmann manifold (G-manifold). Experimental results show that classifier ensembles and their cascades built using R-manifolds are less dependent on some properties of individual classifiers (e.g. depth of decision trees in random forests (RFs) or extra trees (ETs)) in comparison to G-manifolds and Euclidean geometry. More independent individual classifiers allow obtaining R-manifolds with better properties for classification. Generally, the accuracy of classification in nonlinear geometry is higher than in Euclidean one. For multi-class problems, G-manifolds perform similarly to stacking-based classifiers built on R-manifolds of SPD matrices in terms of classification accuracy.

In vision-driven Intelligent Transportation Systems (ITS) where cameras play a vital role, accurate detection and re-identification of vehicles are fundamental demands. Hence, recent approaches have employed a wide range of algorithms to provide the best possible accuracy. These methods commonly generate a vehicle detection model based on its visual appearance features such as license plate, headlights, or some other distinguishable specifications. Among different object detection approaches, Deep Neural Networks (DNNs) have the advantage of magnificent detection accuracy in case a huge amount of training data is provided. In this paper, a robust approach for license plate detection (LPD) based on YOLO v.3 is proposed which takes advantage of high detection accuracy and real-time performance. The mentioned approach can detect the license plate location of vehicles as a general representation of vehicle presence in images. To train the model, a dataset of vehicle images with Iranian license plates has been generated by the authors and augmented to provide a wider range of data for test and train purposes. It should be mentioned that the proposed method can detect the license plate area as an indicator of vehicle presence with no Optical Character Recognition (OCR) algorithm to distinguish characters inside the license plate. Experimental results have shown the high performance of the system with a precision 0.979 and recall 0.972.

Automatic analysis of facial beauty has become an emerging computer vision problem in recent years. Facial beauty prediction (FBP) aims at developing a human-like model that automatically makes facial attractiveness predictions. In this study, we present and evaluate a face attractiveness prediction approach using facial parts as well as a multi-task learning scheme. First, a deep convolutional neural network (CNN) pre-trained on massive face datasets is utilized for face attractiveness prediction, which is capable of automatic learning of high-level face representations. Next, we extend our deep model to other facial attribute recognition tasks. Hence, a multi-task learning scheme is leveraged by our deep model to learn optimal shared features for three correlated tasks (i.e. facial beauty assessment, gender recognition as well as ethnicity identification). To further enhance the attractiveness computation accuracy, specific regions of face images (i.e. left eye, nose and mouth) as well as the whole face are fed into multi-stream CNNs (i.e. three two-stream networks). Each two-stream network adopts a facial part as well as the full face as input. Extensive experiments are conducted on the SCUT-FBP5500 benchmark dataset, where our approach indicates significant improvement in accuracy over the other state-of-the-art methods.

Peripheral Blood Smear (PBS) analysis is a vital routine test carried out by hematologists to assess some aspects of humans' health status. PBS analysis is prone to human errors and utilizing computer-based analysis can greatly enhance this process in terms of accuracy and cost. Recent approaches in learning algorithms, such as deep learning, are data hungry, but due to the scarcity of labeled medical images, researchers had to find viable alternative solutions to increase the size of available datasets. Synthetic datasets provide a promising solution to data scarcity, however, the complexity of blood smears' natural structure adds an extra layer of challenge to its synthesizing process. In this work, we propose a methodology that utilizes Locality Sensitive Hashing (LSH) to create a novel balanced dataset of 2500 synthetic blood smears. This dataset, which was automatically annotated during the generation phase, will be made public for research purposes and covers 17 essential categories of blood cells. We proved the effectiveness of the proposed dataset by utilizing it for training a deep neural network, this model got a very high accuracy score of 98.72% when tested with the well known ALL-IDB dataset. The dataset also got the approval of 5 experienced hematologists to meet the general standards of making thin blood smears.

The process of detecting vehicles' license plates, along with recognizing the characters inside them, has always been a challenging issue due to various conditions. These conditions include different weather and illumination, inevitable data acquisition noises, and some other challenging scenarios like the demand for real-time performance in state-of-the-art Intelligent Transportation Systems (ITS) applications. This paper proposes a method for vehicle License Plates Detection (LPD) and Character Recognition (CR) as a unified application that presents significant accuracy and real-time performance. The mentioned system is designed for Iranian vehicle license plates, which have the characteristics of different resolution and layouts, scarce digits/characters, various background colors, and different font sizes. In this regard, the system uses a separate fine-tuned You Only Look Once (YOLO) version 3 platform for each of the mentioned phases and extracts Persian characters from input images in two automatic steps. For training and testing stages, a wide range of vehicle images in different challenging and straightforward conditions have been collected from practical systems installed as surveillance applications. Experimental results show an end-to-end accuracy of 95.05% on 5719 images. The test data included both color and grayscale images containing the vehicles with different distances and shooting angles with various brightness and resolution. Additionally, the system can perform the LPD and CR tasks in an average of 119.73 milliseconds for real life data, which illustrates a real-time performance for the system and usable applicability. The system is fully automated, and no pre-processing, calibration or configuration procedures are needed.

In multiclass classification, convolutional neural network (CNN) is generally coupled with the cross-entropy (CE) loss, which only penalizes the predicted probability corresponding to a ground truth class and ignores the interclass relationship. We argue that CNN can be improved by using a better loss function. On the other hand, the Wasserstein distance (WD) is a well-known metric used to measure the distance between two distributions. Directly solving the WD problem requires a prohibitively large amount of computation time, whereas the cheaper iterative algorithms have a variety of shortcomings such as computational instability and difficulty in selecting parameters. In this paper, we address these issues by giving an analytical solution to the WD problem—for the first time, we find that for two distributions in hierarchically organized data space, WD has a closed-form solution, which we call “hierarchical WD (HWD).” We use this theory to construct novel loss functions that overcome the shortcomings of CE loss. To this end, multi-CNN information fusion that provides the basis for building category hierarchies is carried out first. Then, the semantic relationship among classes is modeled as a binary tree. Then, CNN coupled with an HWD-based loss, i.e., hierarchical Wasserstein CNN (HW-CNN), is trained to learn deep features. In this way, prior knowledge about the interclass relationship is embedded into HW-CNN, and information from several CNNs provides guidance in the process of training individual HW-CNNs. We conducted extensive experiments over two publicly available remote sensing data sets and achieved a state-of-the-art performance in scene classification tasks.

Feature selection is very important for pattern recognition to reduce the dimensions of data and to improve the efficiency of learning algorithms. Recent research on new approaches has focused mostly on improving accuracy and reducing computing time. This paper presents a flexible feature-selection method based on an optimized kernel mutual information (OKMI) approach. Mutual information (MI) has been applied successfully in decision trees to rank variables; its aim is to connect class labels with the distribution of experimental data. The use of MI removes irrelevant features and decreases redundant features. However, MI is usually less robust when the data distribution is not centralized. To overcome this problem, we propose to use the OKMI approach, which combines MI and a kernel function. This approach may be used for feature selection with nonlinear models by defining kernels for feature vectors and class-label vectors. By optimizing the objection equations, we develop a new feature-selection algorithm that combines both MI and kernel learning, we discuss the relationship among various kernel-selection methods. Experiments were conducted to compare the new technique applied to various data sets with other methods, and in each case the OKMI approach performs better than the other methods in terms of feature-classification accuracy and computing time. OKMI method solves the problem of computation complexity in the probability of distribution, and avoids this problem by finding the optimal features at very low computational cost. As a result, the OKMI method with the proposed algorithm is effective and robust over a wide range of real applications on expert systems.

This survey paper deals with the structural health monitoring systems on the basis of methodologies involving intelligent techniques. The intelligent techniques are the most popular tools for damage identification in terms of high accuracy, reliable nature and the involvement of low cost. In this critical survey, a thorough analysis of various intelligent techniques is carried out considering the cases involved in civil structures. The importance and utilization of various intelligent tools to be mention as the concept of fuzzy logic, the technique of genetic algorithm, the methodology of neural network techniques, as well as the approaches of hybrid methods for the monitoring of the structural health of civil structures are illustrated in a sequential manner.

The accuracies for many pattern recognition tasks have increased rapidly year by year, achieving or even outperforming human performance. From the perspective of accuracy, pattern recognition seems to be a nearly solved problem. However, once launched in real applications, the high-accuracy pattern recognition systems may become unstable and unreliable due to the lack of robustness in open and changing environments. In this article, we present a comprehensive review of research toward robust pattern recognition from the perspective of breaking three basic and implicit assumptions: closed-world assumption, independent and identically distributed assumption, and clean and big data assumption, which form the foundation of most pattern recognition models. Actually, our brain is robust at learning concepts continually and incrementally, in complex, open, and changing environments, with different contexts, modalities, and tasks, by showing only a few examples, under weak or noisy supervision. These are the major differences between human intelligence and machine intelligence, which are closely related to the above three assumptions. After witnessing the significant progress in accuracy improvement nowadays, this review paper will enable us to analyze the shortcomings and limitations of current methods and identify future research directions for robust pattern recognition.

Detecting a counterfeit coin using 2D image processing is nearly impossible in some cases, especially when the coin is damaged, corroded or worn out. Edge detection is one of the most widely used techniques to extract features from 2D images. However, in 2D images, the height information is missing, losing the hidden characteristics. In this paper, we propose a 3D approach to detect and analyze the precipice borders from the coin surface and extract significant features to train an ensemble classification system. To extract the features, we also propose Binned Borders in Spherical Coordinates (BBSC) to analyze different parts of precipice borders at different polar and azimuthal angles. The proposed method is robust even against degradation which appears on shiny coins after 3D scanning. Therefore, there is no need to restore the degraded images before the feature extraction process. Here, the system has been trained and tested with four types of Danish and two types of Chinese coins. We take advantage of stack generalization to classify the coins and add the reject option to increase the reliability of the system. The results illustrate that the proposed method outperforms other counterfeit coin detectors. The accuracy obtained by testing Danish 1990, 1991, 1996, and 2008 datasets are 98.6%, 98.0%, 99.8%, and 99.9% respectively. In addition, results for half Yuan Chinese 1942 and one Yuan Chinese 1997 were 95.5% and 92.2% respectively.

Fake coins are harmful for society, the detection of which is of paramount importance. Due to the large quantities of fake coins in the real world, it is impossible to examine them manually. To address this issue, we present an intelligent system to automatically detect fake coins based on their images. The intelligent system consists of two components: coin image representation and classifier learning. To represent the coin image, a new spatially enhanced bag-of-visual-words model, called SEBOVW model, is proposed. Afterwards, we improve the representation by building a genuine difference subspace. The coin is finally represented based on its projection onto this subspace. In order to discriminate between genuine and fake coins, we train a classifier using the subspace representations. A thorough evaluation of the proposed intelligent system has been conducted on four coin datasets, consisting of thousands of coins of different denominations and from two countries. Promising experimental results in excess of 98 % accuracy demonstrate its effectiveness and validity.

In this paper, we initiate a theoretical study on ${N}$ -expert fusion ( ${N}~\boldsymbol {\geq }~2$ ) in the context of biometric authentication (BA). Optimal fusion weights, which depend on performances and variances of, and correlations among individual base-experts have been found, and we also give and prove some new theorems that serve as the basis for analyzing the performance of the overall system. Our conclusion is that provided that optimal weights are used as fusion coefficients, linear fusion will definitely lead to a better performance than the best individual expert. This contradicts many existing conclusions, which assert that fusion is not always beneficial and that performance improvement due to fusion is guaranteed only when some conditions as to base-experts’ performances, variances, and correlations are satisfied. Besides, for the first time the definition of correlation in the context of BA is clearly and explicitly given to avoid the long-standing ambiguity and vagueness concerning this term, and we make an initial attempt to propose and investigate three types of correlation coefficients. Furthermore, the connection between our proposed optimal fusion method and Fisher’s discriminant is discussed. Extensive experiments have been conducted to confirm our theoretical results and construct counter-examples for the existing conclusions.

Classification finds wide applications in artificial intelligence and expert systems. Feature extraction is a key step for classifier learning. However, the relation among samples is usually ignored in classical feature extraction models. Recently, feature extraction based on graph signal processing that makes use of the relation among samples has attracted great attention. It is a common assumption that the classification information is smooth and of low frequency in these studies. We point out that it is the discrimination ability that essentially makes a good classification feature instead of smoothness. This new perspective prompts us to introduce the concept of discriminative graph signal, and then, based on this concept, we propose a novel feature extraction model for supervised classification. To improve the classification ability for multi-class problems, a generalized model is proposed to extract multiple discriminative signals and an algorithm is also presented to compute the multiple discriminative signals simultaneously. On five publicly available UCI datasets, our proposed method outperforms the existing methods in terms of performance. Finally some drawbacks are discussed and future research directions are also provided.

In the field of biometric authentication, it is a promising trend to perform score fusion to improve authentication accuracy. Many empirical studies have shown the effectiveness of score fusion; however, some other researchers assert that fusion is not always beneficial. Despite considerable empirical efforts, to the best of our knowledge, the research devoted to the theoretical analysis of fusion can be found only in the paper by Poh and Bengio published in 2005. Unfortunately, we find that the variance reduction-equal error rate (VR-EER) model, which is the theoretical basis of this reference, is incorrect and the resulting conclusions are arguable. Besides, we find that the conclusions from several other empirical studies are arguable too. In this paper, using Fermat's theorem and the connection between F-ratio and EER, we conduct a systematic theoretical study on how correlation and performances of base-experts affect fusion, giving the underlying reason why VR-EER model and the above conclusions are wrong. Contrary to these existing conclusions, we prove that provided fusion weights are selected according to our proposed criterion, the combined system will definitely be superior to all the base-experts, regardless of correlation, performances, or variances of base-experts. Experiments are carried out to validate the conclusions of ours and construct counter-examples for the existing conclusions.

Matching pedestrians across disjoint camera views is a challenging task, since their observations are separated in time and space and their appearances may vary considerably. Recently, some approaches of matching pedestrians have been proposed. However, these approaches either used too complex representations or only considered the color information and discarded the spatial structural information of the pedestrian. In order to describe the spatial structural information in color space, we propose a distance-based local binary pattern (DLBP) descriptor. Besides the spatial structural information, the color itself namely its intensity value is also an important feature in matching pedestrians across disjoint camera views. In order to effectively combine these two kinds of information, we further propose a novel CI_DLBP descriptor, which unifies the color intensity and DLBP by learning the joint distributions (2-D histograms) of the DLBP and color intensity at each channel. In addition, different from the previous approaches in which the pedestrians matching is based on their whole bodies, we develop a part-based pedestrian representation because the color density and spatial structural information between the upper outer garment and the lower garment worn by the pedestrian is usually different. Experimental results on challenging realistic scenarios and VIPeR dataset validate the proposed DLBP operator, the CI_DLBP descriptor, and the part-based pedestrian representation for pedestrian matching across disjoint camera views. Compared with existing methods based on color information, this new CI_DLBP approach performs better.

We present algorithms for iris segmentation, feature extraction and selection, and iris pattern matching. To segment the inner boundary from a nonideal iris image, we apply a level set based curve evolution approach using the edge stopping function, and to detect the outer boundary, we employ the curve evolution approach using the regularized Mumford-Shah segmentation model with an energy minimization algorithm. Daubechies wavelet transform (DBWT) is used to extract the textural features, and genetic algorithms (GAs) are deployed to select the subset of informative features by combining the valuable outcomes from the multiple feature selection criteria without compromising the recognition accuracy. To speed up the matching process and to control the misclassification error, we apply a combined approach called the adaptive asymmetrical support vector machines (AASVMs). The parameter values of SVMs are also optimized in order to improve the overall generalization performance. The verification and identification performance of the proposed scheme is validated using the UBIRIS Version 2, the ICE 2005, and the WVU datasets.

In document recognition, it is often important to obtain high accuracy or reliability and to reject patterns that cannot be classified with high confidence. This is the case for applications such as the processing of financial documents in which errors can be very costly and therefore far less tolerable than rejections. This paper presents a new approach based on Linear Discriminant Analysis (LDA) to reject less reliable classifier outputs. To implement the rejection, which can be considered a two-class problem of accepting the classification result or otherwise, an LDA-based measurement is used to determine a new rejection threshold. This measurement (LDAM) is designed to take into consideration the confidence values of the classifier outputs and the relations between them, and it represents a more comprehensive measurement than traditional rejection measurements such as First Rank Measurement and First Two Ranks Measurement. Experiments are conducted on the CENPARMI database of numerals, the CENPARMI Arabic Isolated Numerals Database, and the numerals in the NIST Special Database 19. The results show that LDAM is more effective, and it can achieve a higher reliability while maintaining a high recognition rate on these databases of very different origins and sizes.

A novel and efficient method for correction of slant angles in handwritten numeral strings is proposed. For the first time, the statistical distribution of slant angles in handwritten numerals is investigated and the effects of slant correction on the segmentation of handwritten numeral strings are shown. In our proposed slant correction method, utilizing geometric features, a Component Slant Angle (CSA) is estimated for each connected component independently. A weighted average is then used to compute the String Slant Angle (SSA), which is applied uniformly to correct the slant of all the components in numeral strings. Our experimental results have revealed novel statistics for slant angles of handwritten numeral strings, and also showed that slant correction can significantly improve extraction of segmentation features and segmentation accuracy of touching numerals. Comparison between our slant correction algorithm and similar algorithms in the literature show that our algorithm is more efficient, and on average it has a faster running time.

This paper presents a multimodal approach for a biometrics verification system. It is based on face and hand images captured by a cell phone. The algorithm includes all parts that are required for face and hand verification, such as feature extraction, classification and authentication. To find local facial features, such as eyes, mouth and nose, we apply a point distribution model and active shape models. We use the same system to find distinctive points in hand geometry. The face feature vector is constructed by applying a Gabor filter to the image and extracting the key points found by an active shape model. The palm feature vector contains characteristics of the hand geometry features. A support vector machine (SVM) is applied to verify the identity of the user. One SVM machine is built for each person in the database to distinguish that person from others. To test the algorithm we built our own database containing face and hand images taken by a cell phone camera. The database contains 480 frontal face images and 120 hand images of 30 persons (16 face images and 4 hand images per person).

We propose a variable-length signature for near-duplicate image matching in this paper. An image is represented by a signature, the length of which varies with respect to the number of patches in the image. A new visual descriptor, viz., probabilistic center-symmetric local binary pattern, is proposed to characterize the appearance of each image patch. Beyond each individual patch, the spatial relationships among the patches are captured. In order to compute the similarity between two images, we utilize the earth mover’s distance which is good at handling variable-length signatures. The proposed image signature is evaluated in two different applications, i.e., near-duplicate document image retrieval and near-duplicate natural image detection. The promising experimental results demonstrate the validity and effectiveness of the proposed variable-length signature.

The main objective of Project PalmPrints is to develop and demonstrate a special co-evolutionary genetic algorithm (GA) that optimizes (a clustering fitness function) with respect to three quantities, (a) the dimensions of the clustering space; (b) the number of clusters; and (c) and the locations of the various clusters. This genetic algorithm is applied to the specific practical problem of hand image clustering, with success. In addition to the above, this research effort makes the following contributions: (i) a CD database of (raw and processed) right-hand images; (ii) a number of novel features designed specifically for hand image classification; (iii) an extended fitness function, which is particularly suited to a dynamic (i.e. dimensionality varying) clustering space. Despite the complexity of the multi-optimizational task, the results of this study are clear. The GA succeeded in achieving a maximum fitness value of 99.1%; while reducing the number of dimensions (features) of the space by more than half (from 84 to 41).

Previous research efforts on optical font recognition have mostly limited applications since they deal with only a few types of font attributes and estimate them from a line or block of text. This paper proposes a word-level optical font recognition system for printed Korean and English documents. At the word-level, it has the advantages of obtaining more detailed font attributes including the following: script (Korean and English), font style (regular, bold, italic, and underlined), typeface (Myung-jo and Gothic), point size (10, 12, 14 pts), and word length (2, 3, 4, 5 for Korean, and 4 to 10 for English). A hierarchical classifier and several typographical features have been devised for the system, and their effectiveness are proven by an experiment with a database of 100 sets of 264 font categories.

This paper presents a handwriting recognition system that deals with unconstrained handwriting and large vocabularies. The system is based on the segmentation-recognition paradigm where words are first loosely segmented into characters or pseudocharacters and the final segmentation is obtained during the recognition process, which is carried out with a lexicon. Characters are modeled by multiple hidden Markov models (HMMs), which are concatenated to build up word models. The lexicon is organized as a tree structure, and during the decoding words with similar prefixes share the same computation steps. To avoid an explosion of the search space due to the presence of multiple character models, a lexicon-driven level building algorithm (LDLBA) is used to decode the lexical tree and to choose at each level the more likely models. Bigram probabilities related to the variation of writing styles within the words are inserted between the levels of the LDLBA to improve the recognition accuracy. To further speed up the recognition process, some constraints are added to limit the search efforts to the more likely parts of the search space. Experimental results on a dataset of 4674 unconstrained words show that the proposed recognition system achieves recognition rates from 98% for a 10-word vocabulary to 71% for a 30,000-word vocabulary and recognition times from 9 ms to 18.4 s, respectively.

In this paper a methodology for feature selection for the handwritten digit string recognition is proposed. Its novelty lies in the use of a multiobjective genetic algorithm where sensitivity analysis and neural network are employed to allow the use of a representative database to evaluate fitness and the use of a validation database to identify the subsets of selected features that provide a good generalization. Some advantages of this approach include the ability to accommodate multiple criteria such as number of features and accuracy of the classifier, as well as the capacity to deal with huge databases in order to adequately represent the pattern recognition problem. Comprehensive experiments on the NIST SD19 demonstrate the feasibility of the proposed methodology.

A fast support vector machine (SVM) training algorithm is proposed under SVM's decomposition framework by effectively integrating kernel caching, digest and shrinking policies and stopping conditions. Kernel caching plays a key role in reducing the number of kernel evaluations by maximal reusage of cached kernel elements. Extensive experiments have been conducted on a large handwritten digit database MNIST to show that the proposed algorithm is much faster than Keerthi et al.'s improved SMO, about nine times. Combined with principal component analysis, the total training for ten one-against-the-rest classifiers on MNIST took less than an hour. Moreover, the proposed fast algorithm speeds up SVM training without sacrificing the generalization performance. The 0.6% error rate on MNIST test set has been achieved. The promising scalability of the proposed scheme paves a new way to solve more large-scale learning problems in other domains such as data mining.

In this paper, a two-stage HMM-based recognition method allows us to compensate for the possible loss in terms of recognition performance caused by the necessary trade-off between segmentation and recognition in an implicit segmentation-based strategy. The first stage consists of an implicit segmentation process that takes into account some contextual information to provide multiple segmentation-recognition hypotheses for a given preprocessed string. These hypotheses are verified and re-ranked in a second stage by using an isolated digit classifier. This method enables the use of two sets of features and numeral models: one taking into account both the segmentation and recognition aspects in an implicit segmentation-based strategy, and the other considering just the recognition aspects of isolated digits. These two stages have been shown to be complementary, in the sense that the verification stage compensates for the loss in terms of recognition performance brought about by the necessary tradeoff between segmentation and recognition carried out in the first stage. The experiments on 12,802 handwritten numeral strings of different lengths have shown that the use of a two-stage recognition strategy is a promising idea. The verification stage brought about an average improvement of 9.9% on the string recognition rates. On touching digit pairs, the method achieved a recognition rate of 89.6%.

This paper describes an e0ort towards the development of Arabic cheque databases for research in the recognition of hand-written Arabic cheques. Databases of real-life Arabic legal amounts, Arabic sub-words, courtesy amounts, Indian digits, and Arabic cheques are described. This paper highlights some characteristics of the Arabic language and presents the various steps that have been completed to build these databases including segmentation, binarization and data tagging. It also describes a solid validation procedure including grammars and algorithms used to verify the correctness of the tagging process. Detailed descriptions of the database organization and class distribution are included. These databases aim to facilitate experimental comparisons between various recognition method s, andwill be provid edto all interestedresearchers upon request to

A modular system to recognize handwritten numerical strings is proposed. It uses a segmentation-based recognition approach and a recognition and verification strategy. The approach combines the outputs from different levels such as segmentation, recognition, and postprocessing in a probabilistic model. A new verification scheme which contains two verifiers to deal with the problems of oversegmentation and undersegmentation is presented. A new feature set is also introduced to feed the oversegmentation verifier. A postprocessor based on a deterministic automaton is used and the global decision module makes an accept/reject decision. Finally, experimental results on two databases are presented: numerical amounts on Brazilian bank checks and NIST SD19. The latter aims at validating the concept of modular system and showing the robustness of the system using a well-known database.

This paper proposes a general local learning framework to effectively alleviate the complexities of classifier design by means of “divide and conquer” principle and ensemble method. The learning framework consists of a quantization layer which uses generalized learning vector quantization (GLVQ) and an ensemble layer which uses multi-layer perceptrons (MLP). The proposed method is tested on public handwritten character data sets, which obtains a promising performance consistently. In contrast to other methods, the proposed method is especially suitable for a large-scale real-world classification problems although it is easily scaled to a small training set while preserving a good performance.

The automation of business form process- ing is attracting intensive research interests due to its wide application and its reduction of the heavy work- load due to manual processing. Preparing clean and clear images for the recognition engines is often taken for granted as a trivial task that requires little atten- tion. In reality, handwritten data usually touch or cross the preprinted form frames and texts, creating tremen- dous problems for the recognition engines. In this pa- per, we contribute answers to two questions: "Why do we need cleaning and enhancement procedures in form processing systems?" and "How can we clean and en- hance the hand-filled items with easy implementation and high processing speed?" Here, we propose a generic system including only cleaning and enhancing phases. In the cleaning phase, the system registers a template to the input form by aligning corresponding landmarks. A unified morphological scheme is proposed to remove the form frames and restore the broken handwriting from gray or binary images. When the handwriting is found touching or crossing preprinted texts, morphological op- erations based on statistical features are used to clean it. In applications where a black-and-white scanning mode is adopted, handwriting may contain broken or hollow strokes due to improper thresholding parameters. There- fore, we have designed a module to enhance the image quality based on morphological operations. Subjective and objective evaluations have been studied to show the effectiveness of the proposed procedures.

Global gray-level thresholding techniques such as Otsu's method, and local gray-level thresholding techniques such as edge-based segmentation or the adaptive thresholding method are powerful in extracting character objects from simple or slowly varying backgrounds. However, they are found to be insufficient when the backgrounds include sharply varying contours or fonts in different sizes. A stroke-model is proposed to depict the local features of character objects as double-edges in a predefined size. This model enables us to detect thin connected components selectively, while ignoring relatively large backgrounds that appear complex. Meanwhile, since the stroke width restriction is fully factored in, the proposed technique can be used to extract characters in predefined font sizes. To process large volumes of documents efficiently, a hybrid method is proposed for character extraction from various backgrounds. Using the measurement of class separability to differentiate images with simple backgrounds from those with complex backgrounds, the hybrid method can process documents with different backgrounds by applying the appropriate methods. Experiments on extracting handwriting from a check image, as well as machine-printed characters from scene images demonstrate the effectiveness of the proposed model.

The generically used expert scheduling system (GUESS) intelligent scheduling toolkit has been built and applied in various scheduling domains. Previously, GUESS included three major scheduling approaches: a heuristic-based (suggestion tabulator) approach, a hill-climbing algorithm, and a genetic algorithm approach. GUESS has now been expanded to include a neural network scheduling approach. This paper discusses the development, implementation and testing results of the neural network scheduling method within GUESS.

This paper describes a novel method for edge feature detection of document images based on wavelet decomposition and reconstruction. By applying the wavelet decomposition technique, a document image becomes a wavelet representation, i.e. the image is decomposed into a set of wavelet approximation coefficients and wavelet detail coefficients. Discarding wavelet approximation, the edge extraction is implemented by means of the wavelet reconstruction technique. In consideration of the mutual frequency, overlapping will occur between wavelet approximation and wavelet details, a multiresolution-edge extraction with respect to an iterative reconstruction procedure is developed to ameliorate the quality of the reconstructed edges in this case. A novel combination of this multiresolution-edge results in clear final edges of the document images. This multi-resolution reconstruction procedure follows a coarser-to-finer searching strategy. The edge feature extraction is accompanied by an energy distribution estimation from which the levels of wavelet decomposition are adaptively controlled. Compared with the scheme of wavelet transform, our method does not incur any redundant operation. Therefore, the computational time and the memory requirement are less than those in wavelet transform.

A combining approach has been studied previously to integrate different parts of handwritten characters for their analysis and recognition. Perfect combinations, by which the characters can be identified with certainty, are important to pattern analysis and character recognition. However, a large number of possible combinations (e.g., 63 combinations for a character partitioned into six parts), also produce a lot of perfect combinations. Hence, it is necessary to determine which of them are most important. In this paper, we propose a methodology of finding the basic crucial combinations, and algorithms to compute them. Compared with perfect combinations, such basic crucial combinations are most significant to the character distinctiveness. Similarly, the largest confusion regions are also identified. Experimental studies have also been conducted using the 89 most frequently used patterns of 36 alphanumeric handprints, to obtain their basic crucial combinations and largest confusion combinations. The results indicate that the ratio of the number of basic crucial combinations to perfect combinations is only 12.6%, and the ratio of the number of the largest confusion regions to the total confusion combinations is 15.6%.

This paper presents a technique for extracting the user-entered information from bankcheck images based on a layout-driven item extraction method. The baselines of checks are detected and eliminated by using gray-level mathematical morphology. A priori information about the positions of data is integrated into a combination of top-down and bottom-up analyses of check images. The handwritten information is extracted by a local thresholding technique and the information lost during baseline elimination is restored by mathematical morphology with dynamic kernels. A goal-directed evaluation of the extraction approaches is proposed, and both qualitative and quantitative analyses show noticeable advantages of the proposed approach over the existing approaches.

This paper describes a new kind of neural network – Quantum Neural Network (QNN) – and its application to the recognition of handwritten numerals. QNN combines the advantages of neural modelling and fuzzy theoretic principles. Novel experiments have been designed for in-depth studies of applying the QNN to both real data and confusing images synthesized by morphing. Tests on synthesized data examine QNN's fuzzy decision boundary with the intention to illustrate its mechanism and characteristics, while studies on real data prove its great potential as a handwritten numeral classifier and the special role it plays in multi-expert systems. An effective decision-fusion system is proposed and a high reliability of 99.10% has been achieved.

This paper describes techniques for stroke extraction used in the recognition of handwritten Chinese characters. A new set of feature points is proposed for the analysis of skeleton images. Based on a geometrical graph, a novel criterion is proposed for the identification of fork points in a skeleton image which correspond to joint points in the original character image. Experimental results indicate that the proposed method correctly determines the fork points, and is effective in unifying the joint points.

In this paper, we propose a new approach to combine multiple features in handwriting recognition based on two ideas: feature selection-based combination and class dependent features. A nonparametric method is used for feature evaluation, and the first part of this paper is devoted to the evaluation of features in terms of their class separation and recognition capabilities. In the second part, multiple feature vectors are combined to produce a new feature vector. Based on the fact that a feature has different discriminating powers for different classes, a new scheme of selecting and combining class-dependent features is proposed. In this scheme, a class is considered to have its own optimal feature vector for discriminating itself from the other classes. Using an architecture of modular neural networks as the classifier, a series of experiments were conducted on unconstrained handwritten numerals. The results indicate that the selected features are effective in separating pattern classes and the new feature vector derived from a combination of two types of such features further improves the recognition rate.

Describes a hidden Markov model-based approach designed to recognize off-line unconstrained handwritten words for large vocabularies. After preprocessing, a word image is segmented into letters or pseudoletters and represented by two feature sequences of equal length, each consisting of an alternating sequence of shape-symbols and segmentation-symbols, which are both explicitly modeled. The word model is made up of the concatenation of appropriate letter models consisting of elementary HMMs and an HMM-based interpolation technique is used to optimally combine the two feature sets. Two rejection mechanisms are considered depending on whether or not the word image is guaranteed to belong to the lexicon. Experiments carried out on real-life data show that the proposed approach can be successfully used for handwritten word recognition.

We introduce the new concept of "veinerization", which produces a graph that contains all the "topological" information needed to derive a wide variety of skeletons. Theoretically, the main contribution is to provide a homogeneous framework for integration of the major concepts described in other related works on digital skeletonization. In practice, the new aspect of this approach is to provide the user with different criteria for selecting the most suitable skeleton for a given application, e.g., the user can select a suitable threshold for obtaining the desirable balance between " having a skeleton without noisy prunes" and "having a skeleton that reflects the initial shape". This algorithm has been tested on numerous kinds of patterns, including pathological ones like fractal sets well-known for the complexity of their shapes.

This paper proposes a novel method which enables a Chinese character recognition system to obtain reliable recognition. In this method, two thresholds, i.e. class region thresholdRk and disambiguity thresholdAk, are used by each Chinese character k when the classifier is designed based on the nearest neighbor rule, where Rk defines the pattern distribution region of character k, and Ak prevents the samples not belonging to character k from being ambiguously recognized as character k. A novel algorithm to derive the appropriate thresholds Ak and Rk is developed so that a better recognition reliability can be obtained through iterative learning. Experiments performed on the ITRI printed Chinese character database have achieved highly reliable recognition performance (such as 0.999 reliability with a 95.14% recognition rate), which shows the feasibility and effectiveness of the proposed method.

This paper includes a description of 3 affiliated oriental languages: Chinese, Japanese, and Korean. It includes a description of the origins of these 3 languages and the inter-relationship among them. Drawn from the viewpoints of several experienced researchers in the field of OCR (Optical Character Recognition) and computational linguistics, it attempts to bring out the intriguing aspects of these 3 ideographic languages, including the formation and composition of pictograms, special features, learning, understanding, contextual information, and recognition of characters and words, and their relations to poetic expressions and pattern recognition techniques. Numerous references are given and comments on future trends are also presented.

This paper presents a model for reading cursive scripts which has an architecture inspired by the behavior of human reading and perceptual concepts. The scope of this study is limited to offline recognition of isolated cursive words. First, this paper describes McClelland and Rumelhart's reading model, which formed the basis of the system. The method's behavior is presented, followed by the main original contributions of our model which are: the development of a new technique for baseline extraction, an architecture based on the chosen reading model (hierarchical, parallel, with local representation and interactive activation mechanism), the use of significant perceptual features in word recognition such as ascenders and descenders, the creation of a fuzzy position concept dealing with the uncertainty of the location of features and letters, and the adaptability of the model to words of different lengths and languages. After a description of our model, new results are presented.