Informatics Publishing Limited

R. Dineshkumar ¹, J. Suganthi ²

Affiliations
1 Anna University, Chennai, Tamil Nadu, IN
2 Hindusthan College of Engineering and Technology, Coimbatore, Tamil Nadu, IN

Abstract

In the fast moving world with the amazingly growing technology, character recognitions play a wide role by providing more scope to perform research in OCR techniques. Sanskrit handwritten recognition has been one of the challenging research areas in the field of pattern recognition. Character recognition is the electronic translation of scanned images of handwritten or printed text into a machine encoded text. The character recognition is a standout amongst the most generally utilized biometric attributes for authentication of persons and document. In this paper proposed an off line handwritten character recognition framework utilizing feed forward neural network. A handwritten Sanskrit character is resized into 20x30 pixels and this character is used for training the neural network. After the training process, the same character is given as input to the neural network with different set of neurons in hidden layer and their recognition accuracy rate for different Sanskrit characters has been calculated and compared. The results of the proposed system yields good recognition accuracy rates comparable to that of other handwritten character recognition systems.

Keywords

Classification, Feed forward Neural Network, Handwritten Sanskrit Character Recognition, Image Extraction, Pre-Processing.

Full Text

   

A. Renugadevi ¹, J. Suganthi ¹, K. Sasi Kala Rani ¹

Affiliations
1 Department of Computer Science and Engineering, Hindusthan College of Engineering and Technology, IN

Abstract

In an optical network, the fiber optic cable is used for communication between the nodes in a network by passing lights. The main problem in optical network is finding the link disjoints as well as optimal solution for the disjoints. To tolerate a single link failure in the network, the enhanced active path first algorithm is used which computes the re-routed back-up path. The multiple link failure in a network called fibre span disjoint path problem is solved using integer linear programming algorithm. The loop back recovery is used to provide pre-planned recovery of link or node failures in a network which allows dynamic choice of routes over pre-planned directions. Considering reliability in a mesh networks, the reliability algorithm helps to achieve the maximum reliability in two-path protection. It addresses the multiple disjoint failures that arise in a network and discusses the best solution between paths shared nodes or links. The unified algorithm is used to generate the optimal results with minimum cost for multiple link failures. The heuristic algorithm namely maximum arbitrary double-link protection algorithm helps to pre-compute the back-up path for double-link failures. In all the above approaches the shortest optimized path must be improved. To find the best shortest path, link-disjoint lightpath algorithm is designed to compute the disjoint occurred in a network and it also satisfies the wavelength continuity constraint in wavelength division multiplexing. A polynomial time algorithm Wavelength Division Multiplexing - Passive Optical Networking is used to compute the disjoint happen in the network. The overall time efficiency is analyzed and performance is evaluated through simulations.

Keywords

Wavelength Division Multiplexing, Wavelength Continuity Constraint, Optical Networks, Loop Back Recovery, Double-Link Failure Recovery, Wavelength Division Multiplexing – Passive Optical Networking.

Full Text

     

J. Suganthi ¹, V. Malathi ²

Affiliations
1 Department of CSE, Raja College of Engineering and Technology, Madurai - 625020, Tamil Nadu, IN
2 Anna University Regional Centre, Madurai - 625020, Tamil Nadu, IN

Abstract

Objectives: A reliable and precise classification of tumor types is of great importance and essential for successful Diagnosis and Drug Discovery. Methods: Gene expression profiling has shown a great prospective in the outcome prediction of different types of Cancers, with the innovation of Microarray Technology. Microarray cancer data, which have been organized as samples versus genes fashion, are being exploited for the tissue sample classification. They are also useful for identifying potential gene markers in each subtype of cancer, that helps to analyze a particular cancer type in a successful manner. Findings: In this paper a new method for classification based on Fuzzy Rough-Set Feature Selection Approach through Transductive SVM Technique, which is called as FFS+TSVM is proposed. Moreover recently, for Cancer Pattern Classification, a combined Consistency based Feature Selection Approach through Transductive Support Vector Machine (CBFS+TSVM) has proposed and its prediction accuracy has encouraged than that of existing other Classifiers. However, from the literature survey, it is revealed that the performance of the existing scheme can be improved if Fuzzy Rough Set Approach for Gene Feature Selection with Transductive Support Vector Machine is employed. The present work is implemented with the help of Bio Weka and studied thoroughly in terms of Computational Cost, Dimensionality Reduction, Threshold, Classification Error and Classification Accuracy. Applications/Improvements: The proposed work outperforms the existing Transductive Support Vector Machine (TSVM) in terms of Dimensionality Reduction, Threshold, Classification Error and Classification Accuracy for various Cancer Patterns.

Keywords

Fuzzy Rough Sets, Gene Selection, Information Measures, Low-Density Separation (LDS), Microarray Analysis, Semisupervised Classification, Support Vector Machines (SVM)

Full Text