Unconstrained Handwritten Character Recognition Research Articles

Off-line Odia Handwritten Character Recognition: an Axis Constellation Model Based Research

Handwritten Character Recognition is most challenging area of research, in which for various aspects a little enhancement can be always achieved. It is due to the irregularity of writing and shapes of different class user’s orientation affects the recognition rate. In this paper we have taken the complexity of Odia handwritten character recognition and successfully resolve with Principal Component Analysis (PCA). Here we had adopted a model in which the importance of symmetric axis chords in recognition of unconstrained handwritten characters is established. This symmetric axis chords are drawn along both row-wise and column-wise among the points one end to other. In addition to we have calculated the statistical feature as Euclidian distance, Hamilton distance which drawn from the midpoint of the symmetric chord to nearest pixel of the character. Apart from it we have also reported the angular values from the centroid of the image to the character pixel. This empirical model also harnessed the PCA over the feature set and perform the dimension reduction to the feature set which later termed as the key feature set. A certain series of experiment was carried on for the proper implementation of proposed technique, henceforth we have taken the standard Handwritten Database from various research institutes. Lastly on simulation analysis Radial Basis Function Neural Network (RBFNN) has been reported as to achieve high recognition rate through Gaussian kernel and a comparison among them has also reported here with.

Recognition of online unconstrained handwritten Gurmukhi characters based on Finite State Automata

This paper presents a language-based efficient post-processing algorithm for the recognition of online unconstrained handwritten Gurmukhi characters. A total of 93 stroke classes have been identified to recognize the Gurmukhi character set in this work. Support Vector Machine (SVM) classifier has been employed for stroke classification. The main objective of this paper is to improve the character level recognition accuracy using an efficient Finite State Automata (FSA)-based formation of Gurmukhi characters algorithm. A database of 21,945 online handwritten Gurmukhi words is primarily collected in this experiment. After analysing the collected database, we have observed that a character can be written using one or more strokes. Therefore, a total of 65,946 strokes have been annotated using the 93 identified stroke classes. Among these strokes, 15,069 stroke samples are considered for training the classifier. The proposed system achieved promising recognition accuracy of 97.3% for Gurmukhi characters, when tested with a new database of 8,200 characters, written by 20 different writers.

Compact MQDF classifiers using sparse coding for handwritten Chinese character recognition

The modified quadratic discriminant function (MQDF) is an effective classifier for handwritten Chinese character recognition (HCCR). However, it suffers from high memory requirement for the storage of its parameters, which makes it impractical to be embedded in memory limited hand-held devices. In this paper, we explore the applicability of sparse coding to build compact MQDF classifiers. To be specific, we use sparse coding to compact the parameters of MQDF. Two methods of sparse coding, viz., the maximum likelihood-based method and the K-SVD method, are adopted to build two compact MQDF classifiers, namely, MQDF-ML classifier and MQDF-KSVD classifier. Furthermore, we learn multiple dictionaries rather than single dictionary for sparse coding, because the multiple dictionary learning is capable of not only greatly reducing the computational complexity, but also alleviating the degradation of recognition accuracy, compared to the single dictionary learning. Experiments and comparison with the existing method have demonstrated the effectiveness of our proposed method for the issue of unconstrained handwritten Chinese character recognition.

Online unconstrained handwritten Tibetan character recognition using statistical recognition method

This paper describes a recognition system for online handwritten Tibetan characters using advanced techniques in character recognition. To eliminate noise points of handwriting trajectories, we introduce a de-noising approach by using dilation, erosion, thinning operators of mathematical morphology. Selecting appropriate structuring elements, we can clear up large amounts of noises in the glyphs of the character. To enhance the recognition performance, we adopt a three-stage classification strategy, where the top rank output classes by the baseline classifier are re-classified by similarcharacter discrimination classifier. Experiments have been carried out on two databases MRG-OHTC and IIP-OHTC. Test results show the used recognition algorithm is effective and can be applied to pen-based mobile devices.

BESAC: Binary External Symmetry Axis Constellation for unconstrained handwritten character recognition

We propose a novel perception driven feature extraction called binary external symmetry axis constellation (BESAC) and a fast Boolean matching character recognition technique. A constellation model using a set of external symmetry axes which are perceptually significant can uniquely represent a handwritten character pattern. This model generates two histograms of orientations that are binary coded and concatenated to produce the proposed BESAC feature. A two stage classification strategy is adopted where a parallel Hamming Distance dissimilarity matching is performed on the extracted BESAC feature to achieve fast recognition along with perceptual closure part detection on look-alike characters. We adopt a 10-fold cross validation strategy to evaluate the performance of our algorithm on two major Indian languages, Bangla and Odia with four benchmark databases (ISI Kolkata Bangla numeral, ISI Kolkata Odia and IITBBS Odia numeral, and a newly created IITBBS Odia character database). The average time for classifying an unknown handwritten character is reported to be significantly less than the existing methods. The average recognition accuracy using the proposed approach is proved to outperform the state-of-the-art accuracy results on ISI Kolkata Odia numeral database (99.35%), IITBBS Odia numeral (98.9%), ISI Kolkata Bangla numeral database (99.48%) and IITBBS Odia character (95.01%) database.

DropSample: A new training method to enhance deep convolutional neural networks for large-scale unconstrained handwritten Chinese character recognition

Inspired by the theory of Leitner׳s learning box from the field of psychology, we propose DropSample, a new method for training deep convolutional neural networks (DCNNs), and apply it to large-scale online handwritten Chinese character recognition (HCCR). According to the principle of DropSample, each training sample is associated with a quota function that is dynamically adjusted on the basis of the classification confidence given by the DCNN softmax output. After a learning iteration, samples with low confidence will have a higher frequency of being selected as training data; in contrast, well-trained and well-recognized samples with very high confidence will have a lower frequency of being involved in the ongoing training and can be gradually eliminated. As a result, the learning process becomes more efficient as it progresses. Furthermore, we investigate the use of domain-specific knowledge to enhance the performance of DCNN by adding a domain knowledge layer before the traditional CNN. By adopting DropSample together with different types of domain-specific knowledge, the accuracy of HCCR can be improved efficiently. Experiments on the CASIA-OLHDWB 1.0, CASIA-OLHWDB 1.1, and ICDAR 2013 online HCCR competition datasets yield outstanding recognition rates of 97.33%, 97.06%, and 97.51% respectively, all of which are significantly better than the previous best results reported in the literature.

Offline Handwritten Arabic Character Recognition Using Features Extracted from Curvelet and Spatial Domains

Arabic character recognition is a challenging problem in several artificial intelligence applications, especially when recognizing connected cursive letters. Another dimension of complexity is that Arabic characters may form various shapes depending on their positions in the word. As a result, unconstrained handwritten Arabic character recognition has not been well explored. In this study, we propose an efficient algorithm for Arabic character recognition. The new algorithm combines features extracted from curvelet and spatial domains. The curvelet domain is multiscale and multidirectional. Therefore, curvelet domain is efficient in representing edges and curves. Meanwhile, the spatial domain preserves original aspects of the characters. This feature vector is then trained using the back propagation neural network for the recognition task. The proposed algorithm is evaluated using a database containing 5,600 handwritten characters from 50 different writers. A promising average success rate of 90.3% has been achieved. Therefore, the proposed algorithm is suitable for the unconstrained handwritten Arabic character recognition applications.

A character image restoration method for unconstrained handwritten Chinese character recognition

Despite the success of methods on constrained handwriting databases, recognition of unconstrained handwritten Chinese characters remains a big challenge. One difficulty for recognizing unconstrained handwritting is that some connected strokes are involved or some strokes are omitted. In this paper, a character image restoration method is proposed for unconstrained handwritten Chinese character recognition. In this method, the observed character image is modeled as the combination of the ideal character image with two types of noise images: the omitted stroke noise image and the added stroke noise image. To preserve the original gradient features, restoration is done on the gradient features. The estimated features are then used to discriminate similar characters. To show the effectiveness of the proposed method, we extend some state-of-the-art classifiers based on the estimated features. Experimental results show that the extended classifiers outperform the original state-of-the-art classifiers. This demonstrates that the estimated features are useful for further improving the recognition rate.

Handwritten Character Recognition Using Depth Information by Kinect

t is quite constrained for us to use some other input devices to communicate with computers. In this paper, we integrate human-computer interaction technologies with handwritten Chinese character recognition strategies using depth image information provided by Kinect sensor to realize an unconstrained handwritten character recognition system, which only uses our hand as input device. We predefine several hand gestures as instructions, and for the recognition of these hand gestures, we calculate the contour and fingertips of the hand used for writing using depth image taken by Kinect. By mimicking the functionalities of the computer mouse only using our hands, we can write freely in the air and get the original character image. After Gaussian blurring and normalization, we adopt some classic handwritten character recognition schemes to accomplish the recognition task. Experiments show that the system gives a good result.

Two-tier architecture for unconstrained handwritten character recognition

In this paper, we propose an approach that combines the unsupervised and supervised learning techniques for unconstrained handwritten numeral recognition. This approach uses the Kohonen self-organizing neural network for data classification in the first stage and the learning vector quantization (LVQ) model in the second stage to improve classification accuracy. The combined architecture performs better than the Kohonen self-organizing map alone. In the proposed approach, the collection of centroids at different phases of training plays a vital role in the performance of the recognition system. Four experiments have been conducted and experimental results show that the collection of centroids in the middle of the training gives high performance in terms of speed and accuracy. The systems developed also resolve the confusion between handwritten numerals.

Unconstrained handwritten character recognition based on fuzzy logic

This paper presents an innovative approach called box method for feature extraction for the recognition of handwritten characters. In this method, the binary image of the character is partitioned into a fixed number of subimages called boxes. The features consist of vector distance ( γ) from each box to a fixed point. To find γ the vector distances of all the pixels, lying in a particular box, from the fixed point are calculated and added up and normalized by the number of pixels within that box. Here, both neural networks and fuzzy logic techniques are used for recognition and recognition rates are found to be around 97 percent using neural networks and 98 percent using fuzzy logic. The methods are independent of font, size and with minor changes in preprocessing, it can be adopted for any language.

A connectionist expert system model for conflict resolution in unconstrained handwritten numeral recognition

The paper describes a neural network and expert system model for conflict resolution of unconstrained handwritten numerals. The basic recognizer is the neural network. It solves most of the cases, but will fails in certain confusing cases. The expert system, the second recognizer, resolves the confusions identified by the neural network. The neural network classifier is a combination of Modified Self-Organizing Map (MSOM) and Learning Vector Quantization (LVQ). The experiments are conducted on Self-Organizing Map (SOM), MSOM, and the combinations of SOM & LVQ and MSOM & LVQ techniques. These experiments on a database of unconstrained handwritten samples show that the combination of MSOM & LVQ achieves satisfactory results in terms of classification, recognition and training time. The results obtained from this two-tier architecture are compared with the comments collected from an experiment conducted with a group of human experts specialized in unconstrained handwritten character recognition. The developed system is found to be useful in resolving conflicts in the recognition of Unconstrained Handwritten Numerals of PIN or ZIP codes of mailing addresses.

Automatic construction of structural models incorporating discontinuous transformations

Presents an approach to automatic construction of structural models incorporating discontinuous transformations, with emphasis on application to unconstrained handwritten character recognition. The author considers this problem as constructing inductively, from the data set, some shape descriptions that tolerate certain types of shape transformations. The approach is based on the exploration of complete, systematic, high-level models on the effects of the transformations, and the generalization process is controlled and supported by the high-level transformation models. An analysis of the a priori effects of commonly occurring discontinuous transformations is carried out completely and systematically, leading to a small, tractable number of distinct cases. Based on this analysis, an algorithm for the inference of super-classes under these transformations is designed. Furthermore, through examples and experiments, the author shows that the proposed algorithm can generalize unconstrained handwritten characters into a small number of classes, and that one class can represent various deformed patterns.

A Structural Model of Curve Deformation by Discontinuous Transformations

Structural deformation caused by discontinuous transformations is an intractable problem in shape analysis and description. Structural descriptions depend on the topological structure of the shape, and therefore, they are sensitive to discontinuous transformations which change the topology of the shape. Because of the difficulties, there have been few systematic studies analyzing and modeling structural deformations caused by discontinuous transformations. In this paper, as a first step for overcoming the difficulties, we give a complete and systematic analysis of structural deformations of curves due to some types of commonly occurring discontinuous transformations, in terms of the curve description based on quasi-convexity/concavity incorporating quantized directional features. The transformation laws obtained by the analysis are composed of a small and tractable number of distinct cases. The analysis is applied to the automatic construction of class descriptions from data. In order to show the practical effectiveness of the analysis, experimental results are given for unconstrained handwritten character recognition using a real and standard data set. High recognition accuracy is attained for a variety of deformed patterns with a small number of prototypes (typically one class for one character). We also mention some other applications of the analysis such as the deduction of possible structural descriptions from the class descriptions and the examination of the status of conflicts among the class descriptions.

An approach to integration of off-line and on-line recognition of handwriting

On-line recognition algorithms free from writing constraints and high-quality thinning algorithms are important subjects in research on handwriting recognition and are also essential for the integration of off-line and on-line recognition of handwriting. We present an approach to the integration of off-line and on-line recognition of unconstrained handwritten characters by adapting an on-line recognition algorithm to off-line recognition, based on high-quality thinning algorithms. In the experiments, high recognition rate has been attained with a small number of class descriptions (typically one class for one character).

Adaptive nonlinear shape matching for unconstrained handwritten character recognition

In this paper, we propose an adaptive nonlinear shape matching method which can compensate for the various distortions in unconstrained handwritten characters. In the proposed method, structural information is incorporated to improve the accuracy of matching, and only neighboring pixels of each black pixel are considered to reduce the computational complexity of single matching procedure. Also, iterative nonlinear shape matching procedures in each subregion are adaptively accomplished according to the results of that subregion, in order to accelerate the convergence speed of the matching procedure. In order to verify the performance of the proposed method, experiments with large-set unconstrained handwritten Hangul character database PE92 have been performed. Experimental results reveal that the proposed method is superior to the previous nonlinear shape matching method in processing speed and accuracy of matching.

AN ADAPTIVE MODULAR NEURAL NETWORK WITH APPLICATION TO UNCONSTRAINED CHARACTER RECOGNITION

The topology and the capacity of a traditional multilayer neural system, as measured by the number of connections in the network, has surprisingly little impact on its generalization ability. This paper presents a new adaptive modular network that offers superior generalization capability. The new network provides significant fault tolerance, quick adaption to novel inputs, and high recognition accuracy. We demonstrate this paradigm on recognition of unconstrained handwritten characters.