Browsing by Author "Thendral, Tharmalingam"

Now showing 1 - 3 of 3

AN EFFICIENT CONVOLUTIONAL NEURAL NETWORK BASED CLASSIFIER TO PREDICT TAMIL WRITER
(Periodicals of Engineering and Natural Sciences, 2018-06) Thendral, Tharmalingam; Vijaya, Vijayakumar
Identification of Tamil handwritten calligraphies at different levels such as character, word and paragraph is complicated when compared to other western language scripts. None of the existing methods provides efficient Tamil handwriting writer identification (THWI). Also offline Tamil handwritten identification at different levels still offers many motivating challenges to researchers. This paper employs a deep learning algorithm for handwriting image classification. Deep learning has its own dimensions to generate new features from a limited set of training dataset. Convolutional Neural Networks (CNNs) is one of deep, feed-forward artificial neural network is applied to THWI. The dataset collection and classification phase of CNN enables data access and automatic feature generation. Since the number of parameters is significantly reduced, training time to THWI is proportionally reduced. Understandably, the CNNs produced much higher identification rate compared with traditional ANN at different levels of handwriting.
A HYBRID LINEAR KERNEL WITH PCA IN SVM PREDICTION MODEL OF TAMIL WRITING PATTERN
(International Journal of Simulation: Systems, Science and Technology, 2018) Thendral, Tharmalingam; Vijaya, Vijayakumar
Principal Component Regression (PCR) is a regression analysis technique based on Principal Component Analysis (PCA) which enables the identification of the principal components that can be used in a linear kernel and Support Vector Machine (SVM) as a classifier. In PCR, instead of regressing the dependent variable on the explanatory variables directly, the principal components of the explanatory variables are used as regresseors. Only a subset of all the principal components is made use of for regression, thus making PCR a kind of regularized procedure. The principal components with higher variances are selected as regressors and used in SVM linear kernel to estimate the coefficients of the kernel and the linear kernel is specified as Principal Component Kernel–Support Vector Machine (PCK-SVM). Writer Identification in Tamil handwriting is implemented by employing PCK-SVM and the results of the PCK-SVM are compared with our Weighted Least Square regression Kernel based Support Vector Machine (WLK-SVM) and Bayesian linear regression Kernel based Support Vector Machine (BLK-SVM)models. These methods are evaluated on several text images of handwriting at character, word and paragraph levels. The results show that modified linear kernel performs very well with minimum time taken to classify the writer. Performance comparison results of three kernels achieved highest performance of 94.9% accuracy in PCK-SVM than in WLK of 90.8% and BLK of 92.3% accuracy.
LINEAR KERNEL WITH WEIGHTED LEAST SQUARE REGRESSION CO-EFFICIENT FOR SVM BASED TAMIL WRITER IDENTIFICATION
(Blue Eyes Intelligence Engineering & Sciences Publication, 2019-07) Thendral, Tharmalingam; Vijaya, Vijayakumar
Tamil writer identification is the task of identifying writer based on their Tamil handwriting. Our earlier work of this research based on SVM implementation with linear, polynomial and RBF kernel showed that linear kernel attains very low accuracy compared to other two kernels. But the observation shows that linear kernel performs faster than the other kernels and also it shows very less computational complexity. Hence, a modified linear kernel is proposed to enrich the performance of the linear kernel in recognizing the Tamil writer. Weighted least square parameter estimation method is used to estimate the weights for the dot products of the linear kernel. SVM implementation with modified linear kernel is carried out on different text images of handwriting at character, word and paragraph levels. Comparing the performance with linear kernel, the modified kernel with weighted least square parameter reported promising results.