Mukesh M. Goswami

Classification of printed Gujarati characters using som based k-Nearest Neighbor Classifier

This paper presents a method for combining Self Organizing Map (SOM) with k-Nearest Neighbor Classifier (k-NN) to device an elegant classification technique and applying it for classification of subset of printed Gujarati characters. Many researchers have employed many different models for the classification of printed/handwritten characters for number of different languages all over the globe; few of the widely used classifiers are Template Matching, Artificial Neural Network (ANN), Hidden Markov Model (HMM), and Support Vector Machine (SVM) etc. Our attempt is to use SOM based k-NN classifier for classification of subset of printed Gujarati characters. This approach does not require prior feature identification stage hence it is faster and more generalize compare to other approaches. A prototype system is implemented for the same and tested on sufficient dataset. Average accuracy of 82.36% is reported on test dataset.

Offline handwritten Gujarati numeral recognition using low-level strokes

This paper focuses on the development of offline handwritten Gujarati numeral database of reasonable size and its recognition using low-level stroke features. The database consists of 14,000 samples collected from 140 people with different age group, educational background, and work culture. A novel technique for the extraction of various low-level stroke features, like endpoints, junction points, line segments, and curve segments, is proposed, and the block-wise histogram of low-level stroke features is used for the recognition of offline handwritten numerals from two of the popular Indian scripts, namely Gujarati and Devanagari. The baseline experiments were performed using k-nearest neighbour (k-NN) classifier, and the results were further improved by using the statistically advance support vector machine (SVM) classifier with radial basis function (RBF) kernel. The average test accuracy obtained on Gujarati and Devanagari database were 98.46% and 98.65%, respectively, which is comparable to other existing work. The experiments were also performed on the mixed numerals recognition from Gujarati-Devanagari and Gujarati-English considering the multi-script scenarios in Indian documents.

Similar looking Gujarati printed character recognition using Locality Preserving Projection and artificial neural networks

An attempt has been made to recognize similar looking printed Gujarati characters. It has been assumed that similar looking character recognition is the main challenge for an OCR to work efficiently. The dimensionality of the character images is drastically and very efficiently reduced and presented by only a few significant coefficients. A new mechanism called ESLPP is introduced for the same. Coefficients extracted using ESLPP which explores the natural grouping, preserves the local structure of the data and finds the essential data manifold structure, are then fed to the Neural Network as features. Six datasets each consisting of two or more similar looking characters are used for experimentation. The recognition accuracy in all the experiments is found to be very satisfactory.

On-line handwritten Gujarati character Recognition using low level stroke

This paper presents a low level stroke feature based method for recognition of online handwritten Gujarati characters and numerals. A reasonable size database of online handwritten Gujarati characters and numerals has been developed. This is the first such database of online handwritten symbols for Gujarati script The hierarchical histograms of twelve different low level stroke features and eight directional features were generated to capture the variation in strokes at different level. Recognition is performed using a nearest neighbor (i.e. K-NN) classifier with k-fold cross validation on the dataset having 4500 samples from 45 different classes (37 characters and 8 numerals). Overall Recognition rates achieved are 95%, 93% and 90% for numerals dataset, characters dataset and combine dataset of numerals and characters respectively.

Classification of handwritten Gujarati numerals

This paper addresses the problem of recognizing handwritten numerals for Gujarati Language. Three methods are presented for feature extraction. One belongs to the spatial domain and other two belongs to the transform domain. In first technique, a new method has been proposed for spatial domain which is based on Freeman chain code. This method obtains the global direction by considering n × n neighbourhood and thus eliminates the noise which occurs due to local direction. In second and third method, 85 dimensional Fourier descriptors and Discrete Cosine Transform coefficients were computed and treated as feature vectors. Comparative analysis has been done for these three methods. These methods are tested with three different classifiers namely K-Nearest Neighbour, Support Vector Machine and Back Propagation Neural Network. Experimental results were evaluated using 10 fold cross validation. The highest recognition rates obtained for full data set of 3000 digits are 85.67%, 93.60% and 93.00% using modified chain code, DFT and DCT respectively.

Structural Feature Based Classification of Printed Gujarati Characters

This paper presents a Structural feature based method for classification of printed Gujarati characters. The ability to provide incremental definition of characters in terms of its native components makes the proposal unique and versatile. It deals with varied sizes, font styles, and stoke widths. The features are validated on subset of machine printed Gujarati characters using a simple rule based classifier and the initial results are encouraging.

Classification of Printed Gujarati Characters Using Low-Level Stroke Features

This article presents an elegant technique for extracting the low-level stroke features, such as endpoints, junction points, line elements, and curve elements, from offline printed text using a template matching approach. The proposed features are used to classify a subset of characters from Gujarati script. The database consists of approximately 16,782 samples of 42 middle-zone symbols from the Gujarati character set collected from three different sources: machine printed books, newspapers, and laser printed documents. The purpose of this division is to add variety in terms of size, font type, style, ink variation, and boundary deformation. The experiments are performed on the database using a k-nearest neighbor (kNN) classifier and results are compared with other widely used structural features, namely Chain Codes (CC), Directional Element Features (DEF), and Histogram of Oriented Gradients (HoG). The results show that the features are quite robust against the variations and give comparable performance with other existing works.

Proposed efficient approach for classification for multi-relational data mining using Bayesian Belief Network

Multi-Relational Data Mining is an active area of research for researchers from last many decades. Relational database is an important source of structure data, hence richest source of knowledge. Most of the commercial and application oriented data uses relational database system in which multiple relations are link through primary key, foreign key relationship. Thus, the field of Multi-Relational Data Mining (MRDM) deals with extraction of information from relational database containing multiple tables related with each other. In order to extract important information or knowledge, it is required to apply Data Mining algorithms on this relational database but most of these algorithms works only on single table. Generating a single table may result in to loss of important information, like relation between tuples. Also it is a not efficient in terms of time and space. In this research, we propose a Probabilistic Graphical Model, namely Bayesian Belief Network (BBN), based approach that considers not only attributes of table but also the relation between tables. The conditional dependencies between tables is derived from Semantic Relationship Graph (SRG) of the relational database. This research also aims, to find relevant attributes from Multi-Relational dataset in order to improve the accuracy.

Performance Analysis of Supervised & Unsupervised Techniques for Brain Tumor Detection and Segmentation from MR Images

Brain tumor detection and segmentation from the magnetic resonance images (MRI) is a difficult task as in the MR brain images, various tissues such as white matter, gray matter, and cerebrospinal fluid have complicated structures that make it difficult to segment the tumor. An automated system for brain tumor detection and segmentation will help the patients for proper treatment planning. Also, it will improve the diagnosis and reduce the diagnostic time. Segmentation of brain tumor MR images is the most difficult task as the tumor varies in terms of size, shape, location, and texture. In this paper, we discuss various supervised and unsupervised techniques for brain tumor detection and segmentation such as K-nearest neighbor (K-NN), K-means clustering, and using morphological operators. We also review the results obtained.

A survey on offline handwritten north Indian script symbol recognition

North Indian scripts are used in 70% of the regions of India. Devanagari, Bangla, Gujarati, Oriya, and Gurumukhi are the major North Indian scripts listed in decreasing order of their usage. There are ample amount of work done on printed symbol recognition for various North Indian scripts, but handwritten symbol recognition still needs further attention. This paper represents the thorough study of the work done so far for handwritten symbol recognition for major North Indian script. Furthermore, the challenges of each script and various feature extraction techniques and classifiers used are also discussed. The paper will provide a proper startup and almost complete scenario of the state of the art in offline handwritten symbol recognition from North Indian script.