Search

In this paper, we propose integrated adaptive sensitivity and robustness terms for the cost function of a convolutional neural network (CNN). The sensitivity term considers the slight variations and high frequency components of the input image samples. It distinguishes between images that look similar but belong to different classes. This regularizer is designed to enhance the between-class distance which is a biological definition for the simple cells of the visual system. On the other hand, the robustness term is used to develop a more stable CNN structure against disturbances and perturbations. The robust term provides better within-class features because it recognizes images that look different but are actually from the same class. The robust term symbolizes the complex cell characteristics of the visual system. The coefficients of the sensitivity and robustness regularization terms are adaptively tuned along with the network parameters using gradient descent. Two optimizers are assigned to tune the parameters: one for tuning the model parameters and the other one to adjust the sensitivity and robustness coefficients. This approach is applied to Korean traditional documents for character classification. The results show better within- and between-class classification ability for highly complex character styles with imbalanced number of samples.

Recent developments in depth sensing technology such as Leap Motion have opened novel directions in Human-Computer-Interaction (HCI) research domain. The sensor extends the way of writing from traditional method to a gesture based writing in the 3D space. The online text written in 3D space over the sensorâs viewing field is different from traditional 2D handwriting in several ways. The 3D handwriting does not consist any stroke information, since all characters are connected by a single stroke. Moreover, non-uniform text styles and jitters during writing in 3D space create additional challenge for the recognition task. Because of these challenges in 3D handwriting, recognition of cursive words is not satisfied using a single classifier. In this paper, we present a lexicon free approach for the recognition of 3D handwritten words in Latin and Devanagari scripts by combining multiple classifiers. The individual recognition systems are computed using Bidirectional Long-Short Term Memory Neural Network (BLSTM-NN) classifier with the help of different features. The combination of multiple classifier is performed by aligning the output word sequence of each classifier using the Recognizer Output Voting Error Reduction (ROVER) framework. Accuracies of 72.25% and 71.86% are recorded using the proposed methodology for Latin and Devanagari scripts, respectively.

We present an off-line handwritten word recognition system.The system is based on extraction of directional features.Arnold transform followed by Hough transform is used to extract directional feature.A multi-class linear SVM is employed to recognize handwritten cursive words.Experiments are done on CENPERMI database of legal amounts written in English. This paper presents a holistic off-line handwriting recognition system based on extraction of directional features which depends on the stroke orientation distribution of cursive word. This stroke orientation distribution is estimated using Arnold transform followed by Hough transform. Besides this feature some other directional shape features are also used to form feature vector. Finally, a multi-class linear SVM is employed to recognize cursive word. Experiments are carried out on CENPARMI database of legal amount written in English and an overall accuracy of 87.19% is achieved. We have also compared our proposed method with the state-of-the-art methods for handwritten character recognition using C-Cube data-set.

In this paper, we propose a new font and size identification method for ultra-low resolution Arabic word images using a stochastic approach. The literature has proved the difficulty for Arabic text recognition systems to treat multi-font and multi-size word images. This is due to the variability induced by some font family, in addition to the inherent difficulties of Arabic writing including cursive representation, overlaps and ligatures. This research work proposes an efficient stochastic approach to tackle the problem of font and size recognition. Our method treats a word image with a fixed-length, overlapping sliding window. Each window is represented with a 102 features whose distribution is captured by Gaussian Mixture Models (GMMs). We present three systems: (1) a font recognition system, (2) a size recognition system and (3) a font and size recognition system. We demonstrate the importance of font identification before recognizing the word images with two multi-font Arabic OCRs (cascading and global). The cascading system is about 23% better than the global multi-font system in terms of word recognition rate on the Arabic Printed Text Image (APTI) database which is freely available to the scientific community.

This work presents a holistic system for the off-line recognition of cursive words based on the extraction of perceptual features by means of convolution with templates of line segments. The method works directly on gray-level images to avoid losing information due to binarization. Features are represented spatially in order to preserve topological information which is well suited for recognition through convolutional neural networks. Moreover, our feature extraction method does not need to detect baselines to find ascenders and descenders. The system has been tested on small and medium vocabulary sizes demonstrating the feasibility of the proposed method.

This paper presents a cursive Arabic text recognition system. The system decomposes the document image into text line images and extracts a set of simple statistical features from a narrow window which is sliding a long that text line. It then injects the resulting feature vectors to the Hidden Markov Model Toolkit (HTK). HTK is a portable toolkit for speech recognition system. The proposed system is applied to a data corpus which includes Arabic text of more than 600 A4-size sheets typewritten in multiple computer-generated fonts.

This paper analyses a handwriting recognition system for offline cursive words based on HMMs. It compares two approaches for transforming offline handwriting available as two-dimensional images into one-dimensional input signals that can be processed by HMMs. In the first approach, a left-right scan of the word is performed resulting in a sequence of feature vectors. In the second approach, a more subtle process attempts to recover the temporal order of the strokes that form words as they were written. This is accomplished by a graph model that generates a set of paths, each path being a possible temporal order of the handwriting. The recognition process then selects the most likely temporal stroke order based on knowledge that has been acquired from a large set of handwriting samples for which the temporal information was available. We show experimentally that such an offline recognition system using the recovered temporal order can achieve recognition performances that are much better than those obtained with the simple left-right order, and that come close to those of an online recognition system. We have been able to assess the ordering quality of handwriting when comparing true ordering and recovered one, and we also analyze the situations where offline and online information differ and what the consequences are on the recognition performances. For these evaluations, we have used about 30,000 words from the IRONOFF database that features both the online signal and offline signal for each word.

This paper presents a new method on off-line recognition of handwritten Arabic script. The method does not require segmentation into characters, and is applied to cursive Arabic script, where ligatures, overlaps and style variation pose challenges to the recognition system. The method trains a single hidden Markov model (HMM) with the structural features extracted from the manuscript words. The HMM is composed of multiple character models where each model represents one letter from the alphabet. The performance of the proposed method is assessed using samples extracted from a historical handwritten manuscript.

This work presents the application of HMM adaptation techniques to the problem of Off-Line Cursive Script Recognition. Instead of training a new model for each writer, one first creates a unique model with a mixed database and then adapts it for each different writer using his own small dataset.\\ Experiments on a publicly available benchmark database show that an adapted system has an accuracy higher than 80\% even when less than 30 word samples are used during adaptation, while a system trained using the data of the single writer only needs at least 200 words in order to achieve the same performance as the adapted models.

Knowledge concerning the structure of some prominent English letters, as well as the structural characteristics between background regions and character components is investigated as a novel approach to cursive script segmentation. First, connected components consisting of more than one character are split into sub-components based on their face-up or face-down background regions. Secondly, the over-segmented sub-components are merged into characters according to the knowledge of character structures and their joining characteristics. The algorithm achieved around 80% correct segmentation on a difficult database of unconstrained handwritten words.

The concept of key characters in a cursively handwritten word image is introduced and a method for extracting the key characters is presented. Key characters capture the unambiguous parts of the cursive words that can be reliably segmented and recognized. We propose a method for lexicon reduction using key characters in conjunction with a word-length estimation.

We present an on-line Korean character recognition method using a Time Delay Neural Network and a Hidden Markov Model. A TDNN is used to estimate a posteriori probabilities for graphemes in a character. An HMM aligns graphemes onto tentative graphemes, which are produced by heuristic segmentation points, using a modified Viterbi algorithm. The HMM contains statistic information, i.e., the frequencies of grapheme pairs and first consonants in the underlying language. We experiment on the proposed method with multi-writer cursive characters in a boxed input mode. Tests with untrained 22,500 on-line cursive characters show a recognition rate of 93.4% and speed of 1.37 s per character.

In this paper, a dynamic programming algorithm which can extract the line segments from the loci of Chinese character in running hand or cursive writing is proposed. This algorithm tries to find the minimum line segments to fit the cursive stroke of character within the prescribed error bound. Three error measures are presented to construct a globally optimal approximation to the given stroke, including sum of deviation, peak deviation and length error. Extensive experiments are demonstrated to illustrate the performance of the proposed algorithm.

This paper proposes an on-line Chinese character recognition method using Adaptive Resonance Theory (ART) based stroke classification. Strokes, primitive components of Chinese characters, are usually warped into a cursive form and classifying them is very difficult. To deal with such cursive strokes, we consider them as a recognition unit and automatically classify them using an ART-2 neural network. The neural network has the advantage of assembling similar patterns together to form classes in a self-organized manner. This stroke classifier contributes to high stroke recognition rate and less recognition time. A database for character recognition also dynamically constructed with generalized character lists, and a new character can be included simply by adding a new sequence to the list. Character recognition is achieved by traversing the Chinese character database with a sequence of recognized strokes and positional relations between the strokes. To verify the performance of the system, we tested it on 1800 basic Chinese characters used daily in Korea, and obtained a good recognition rate of 93.13%. These results suggest that the proposed system is pertinent to be put into practical use.

This paper describes an algorithm for holistic recognition of off-line cursive words using temporal stroke information derived from off-line script. Temporal information is extracted by traversing the strokes without explicit segmentation of the word into constituent characters. The word image is then mapped on to a feature vector matrix of uptrends and downtrends of strokes. This feature vector matrix is compared to prestored feature vector of lexicon entries and ranked accordingly. On a test set of images, the temporal feature extraction rate is 80%. Given the correct set of temporal features, the recognition rate of the holistic classifier is 81% on small lexicons.

We propose an efficient method for on-line recognition of cursive Korean characters. The method uses a set of extended primitive strokes and fuzzy membership functions to extract cursive strokes. Positional relation of two strokes is computed also using fuzzy functions. With a sequence of stroke vectors and their positional relations computed, the method searches the target character on a character database. A test with 17,500 handwritten characters shows recognition rate of 96.3% with speed of 0.3 second per character.

This paper describes a framework for generating variant samples of a handwritten cursive string based on Hollerbach's Oscillation Theory of Handwriting. A single cursive word is analyzed. Significant features of the word are extracted from the velocity profile of the pen movement. These features are used to generate several samples of the same word. The proposed framework for sample generation can be applied only to continuous cursive strings.

This paper presents a new method for cursive script recognition based on the extraction of key letters which consist of the parts of the handwritten text which are singularities of handwriting. By processing the raw input data representing the handwritten text, the singular features: loops, ascenders, and descenders are first defined and determined. Based on them, the key letters are then extracted. This paper focuses on the extraction of potentially meaningful shapes, corresponding to visual letters, in the recognition of cursive handwriting. First the key letters of a cursive word are found, such as, b, f, p, q, F, T, etc., then a possible word based on these letters is selected. In the testing phase, there are 304 words from 74 cheques in our data base written by 40 writers. 276 words are correctly segmented, that is more than 90% suscessful segmentation. The problems leading to the missegmentation are identified for further improvement.