Handwritten Numeral Databases of Indian Scripts and Multistage Recognition of Mixed Numerals
01 Mar 2009-IEEE Transactions on Pattern Analysis and Machine Intelligence (IEEE Computer Society)-Vol. 31, Iss: 3, pp 444-457
TL;DR: P pioneering development of two databases for handwritten numerals of two most popular Indian scripts, a multistage cascaded recognition scheme using wavelet based multiresolution representations and multilayer perceptron classifiers and application for the recognition of mixed handwritten numeral recognition of three Indian scripts Devanagari, Bangla and English.
Abstract: This article primarily concerns the problem of isolated handwritten numeral recognition of major Indian scripts. The principal contributions presented here are (a) pioneering development of two databases for handwritten numerals of two most popular Indian scripts, (b) a multistage cascaded recognition scheme using wavelet based multiresolution representations and multilayer perceptron classifiers and (c) application of (b) for the recognition of mixed handwritten numerals of three Indian scripts Devanagari, Bangla and English. The present databases include respectively 22,556 and 23,392 handwritten isolated numeral samples of Devanagari and Bangla collected from real-life situations and these can be made available free of cost to researchers of other academic Institutions. In the proposed scheme, a numeral is subjected to three multilayer perceptron classifiers corresponding to three coarse-to-fine resolution levels in a cascaded manner. If rejection occurred even at the highest resolution, another multilayer perceptron is used as the final attempt to recognize the input numeral by combining the outputs of three classifiers of the previous stages. This scheme has been extended to the situation when the script of a document is not known a priori or the numerals written on a document belong to different scripts. Handwritten numerals in mixed scripts are frequently found in Indian postal mails and table-form documents.
Citations
More filters
TL;DR: The proposed BHNR with deep LSTM (BNHR-DLSTM) standardizes the composed numeral images first and then utilizes two layers of L STM to characterize singular numerals and indicates agreeable recognition precision and beat other conspicuous existing methods.
Abstract: Handwritten numeral recognition (HNR) has gained much attention in present days as it can be applied in range of applications. Research on unconstrained HNR has shown impressive progress in few scripts but is far behind for Bangla although it is one of the major languages. Bangla contains similar shaped numerals which are difficult to distinguish even in printed form and this makes Bangla HNR (BHNR) a challenging task. Our goal in this study is to build up a superior BHNR framework and consequently explore the profound design of Long Short Term Memory (LSTM) method. LSTM is a variation of Recurrent Neural Network and is effectively used for sequence ordering with its distinct features. This study considered deep architecture of LSTM for better performance. The proposed BHNR with deep LSTM (BNHR-DLSTM) standardizes the composed numeral images first and then utilizes two layers of LSTM to characterize singular numerals. Benchmark dataset with 22000 handwritten numerals having various shapes, sizes and varieties are utilized to examine the proficiency of BNHR-DLSTM. The proposed method indicates agreeable recognition precision and beat other conspicuous existing methods.
12 citations
Cites background or methods from "Handwritten Numeral Databases of In..."
...Test set precision of work [6] is very close to our proposed work, yet it utilized a scaled-up rendition of the first dataset which is actually 10 times bigger than the one utilized as a part of this study....
[...]
...& Chaudhuri [6] Wavelet filter at different resolutions Four MLPs in...
[...]
...Bhattacharya & Chaudhuri [6] proposed a multistage cascaded recognition scheme using MLP based classifier for doing BHNR, where at first, they applied waveletfiltered images having non-similar resolutions for extracting higher level features followed by cascading of three MLPs for classification....
[...]
...From the ISI benchmark dataset [6], 18000 samples are considered for training and 4000 samples are considered for testing....
[...]
...This study uses the benchmark handwritten numeral image dataset kept up by CVPR unit, ISI, Kolkata [6]....
[...]
TL;DR: A stacked ensemble meta-learning approach for customized convolutional neural network is proposed for Marathi handwritten numeral recognition and it overpowers the average ensemble because the weighted and maximum contribution of each pipeline is taken in this approach.
Abstract: Pattern Recognition is the method of mapping the inputs to their respective target classes based on features of data. In this paper a stacked ensemble meta-learning approach for customized convolutional neural network is proposed for Marathi handwritten numeral recognition. Stacked ensemble merges the pre-trained base pipe lines to create a multi-head meta-learning classifier that outputs the final target labels. It overpowers the average ensemble because the weighted and maximum contribution of each pipeline is taken in this approach. The stacked ensemble meta-learning classifier proves to be efficient because the base pipelines, which are already acquainted with output desirable results, are concatenated, instead of averaging, to achieve maximum efficiency. Performance evaluation and analysis have been done on Marathi handwritten numeral dataset, and the experiment results are better than the existing proposed systems.
11 citations
22 Jul 2014
TL;DR: A new feature extraction method based on the non-redundant Stockwell Transform (ST), which takes care of the redundancy as well as computational complexity of original ST is proposed, which outperforms the previous reported classification results.
Abstract: Feature extraction is an important stage which decides the accuracy of any character recognition system. The state-of-the-art feature extraction can be categorized to be either spatial domain based, transform domain based or a hybrid combination of both. We propose a new feature extraction method based on the non-redundant Stockwell Transform (ST), which takes care of the redundancy as well as computational complexity of original ST. We applied the proposed method on Odia numerals with k-Nearest Neighbor (k-NN) classifier, Support Vector Machine (SVM), Multi-Layer Perceptron (MLP) classifier and Modified Quadratic Discriminant Function (MQDF) classifier. The highest recognition accuracy is found to be 98.80% for the Odia numeral database, which outperforms the previous reported classification results.
11 citations
TL;DR: The experimental results show that the proposed scheme outperforms other state-of-the-art approaches in terms of recognition accuracy andMeta-heuristic optimization schemes can facilitate feature learning even with a small amount of training data.
Abstract: In recent years, the non-handcrafted feature extraction methods have gained increasing popularity for solving pattern classification tasks due to their inherent ability to extract robust features and handle outliers. However, the design of such features demands a large set of training data. Meta-heuristic optimization schemes can facilitate feature learning even with a small amount of training data. This paper presents a new feature learning mechanism called multi-objective Jaya convolutional network (MJCN) that attempts to learn meaningful features directly from the images. The proposed scheme, unlike the convolutional neural networks, comprises a convolution layer, a multiplication layer, an activation layer and an optimizer known as multi-objective Jaya optimizer (MJO). The convolution layer searches meaningful patterns in an image through the local neighborhood connections and the multiplication layer projects the convolutional response to a more compact feature space. The weights used in these layers are initialized randomly and MJO is then introduced to optimize the weights. The main objective of MJO is to maximize the inter-class distance and minimize the intra-class variance. The feature vectors are finally derived using the optimized weights. The derived features are finally fed to a set of standard classifiers for recognition of characters. The performance of the proposed model is evaluated on various benchmark datasets, namely, NITR Odia handwritten character, ISI Kolkata Odia numeral, ISI Kolkata Bangla numeral, and MNIST as well as a newly developed dataset NITR Bangla numeral. The experimental results show that the proposed scheme outperforms other state-of-the-art approaches in terms of recognition accuracy.
10 citations
01 Dec 2016
TL;DR: In the proposed system, neural and non-neural approach is presented for classification of different characters for offline Handwritten Character Recognition (HWCR) system.
Abstract: For development of offline Handwritten Character Recognition (HWCR) system, scripts have always posed a difficulty. In the proposed system, we present neural and non-neural approach for classification of different characters. After pre-processing, features are extracted using Chain code histogram and Intersection junction techniques. BPN, KNN & SVM have been used to train and classify the Devnagari and MODI vowels separately. The combinations have been compared for both the scripts on the basis of their recognition rate.
10 citations
Cites background from "Handwritten Numeral Databases of In..."
...Some research groups are working for Devnagari, Bangla, Gurumukhi, Oriya, Urdu and other scripts [3]....
[...]
...Here, the function is only approximated locally and all computation is deferred until classification[3]....
[...]
References
More filters
01 Jan 1998
TL;DR: In this article, a graph transformer network (GTN) is proposed for handwritten character recognition, which can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters.
Abstract: Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day.
42,067 citations
37,017 citations
Book•
16 Jul 1998
TL;DR: Thorough, well-organized, and completely up to date, this book examines all the important aspects of this emerging technology, including the learning process, back-propagation learning, radial-basis function networks, self-organizing systems, modular networks, temporal processing and neurodynamics, and VLSI implementation of neural networks.
Abstract: From the Publisher:
This book represents the most comprehensive treatment available of neural networks from an engineering perspective. Thorough, well-organized, and completely up to date, it examines all the important aspects of this emerging technology, including the learning process, back-propagation learning, radial-basis function networks, self-organizing systems, modular networks, temporal processing and neurodynamics, and VLSI implementation of neural networks. Written in a concise and fluid manner, by a foremost engineering textbook author, to make the material more accessible, this book is ideal for professional engineers and graduate students entering this exciting field. Computer experiments, problems, worked examples, a bibliography, photographs, and illustrations reinforce key concepts.
29,130 citations
TL;DR: In this paper, it is shown that the difference of information between the approximation of a signal at the resolutions 2/sup j+1/ and 2 /sup j/ (where j is an integer) can be extracted by decomposing this signal on a wavelet orthonormal basis of L/sup 2/(R/sup n/), the vector space of measurable, square-integrable n-dimensional functions.
Abstract: Multiresolution representations are effective for analyzing the information content of images. The properties of the operator which approximates a signal at a given resolution were studied. It is shown that the difference of information between the approximation of a signal at the resolutions 2/sup j+1/ and 2/sup j/ (where j is an integer) can be extracted by decomposing this signal on a wavelet orthonormal basis of L/sup 2/(R/sup n/), the vector space of measurable, square-integrable n-dimensional functions. In L/sup 2/(R), a wavelet orthonormal basis is a family of functions which is built by dilating and translating a unique function psi (x). This decomposition defines an orthogonal multiresolution representation called a wavelet representation. It is computed with a pyramidal algorithm based on convolutions with quadrature mirror filters. Wavelet representation lies between the spatial and Fourier domains. For images, the wavelet representation differentiates several spatial orientations. The application of this representation to data compression in image coding, texture discrimination and fractal analysis is discussed. >
20,028 citations
01 Jan 1988
TL;DR: This chapter contains sections titled: The Problem, The Generalized Delta Rule, Simulation Results, Some Further Generalizations, Conclusion.
Abstract: This chapter contains sections titled: The Problem, The Generalized Delta Rule, Simulation Results, Some Further Generalizations, Conclusion
17,604 citations
Additional excerpts
...Ç...
[...]