Suresh Sundaram

Attention-Feedback Based Robust Segmentation of Online Handwritten Isolated Tamil Words

In this article, we propose a lexicon-free, script-dependent approach to segment online handwritten isolated Tamil words into its constituent symbols. Our proposed segmentation strategy comprises two modules, namely the (1) Dominant Overlap Criterion Segmentation (DOCS) module and (2) Attention Feedback Segmentation (AFS) module. Based on a bounding box overlap criterion in the DOCS module, the input word is first segmented into stroke groups. A stroke group may at times correspond to a part of a valid symbol (over-segmentation) or a merger of valid symbols (under-segmentation). Attention on specific features in the AFS module serve in detecting possibly over-segmented or under-segmented stroke groups. Thereafter, feedbacks from the SVM classifier likelihoods and stroke-group based features are considered in modifying the suspected stroke groups to form valid symbols. The proposed scheme is tested on a set of 10000 isolated handwritten words (containing 53,246 Tamil symbols). The results show that the DOCS module achieves a symbol-level segmentation accuracy of 98.1%, which improves to as high as 99.7% after the AFS strategy. This in turn entails a symbol recognition rate of 83.9% (at the DOCS module) and 88.4% (after the AFS module). The resulting word recognition rates at the DOCS and AFS modules are found to be, 50.9% and 64.9% respectively, without any postprocessing.

An Improved Online Tamil Character Recognition Engine Using Post-Processing Methods

We propose script-specific post processing schemes for improving the recognition rate of online Tamil characters. At the first level, features derived at each sample point of the preprocessed character are used to construct a subspace using the 2DPCA algorithm. Recognition of the test sample is performed using a nearest neighbor classifier. Based on the analysis of the confusion matrix, multiple pairs of confused characters are identified. At the second level, we use script specific cues to sort out the ambiguities among the confused characters. This strategy reduces the recognition error among the confused character sets handled, by more than 4%. This approach can be applied irrespective of the nature of the classifier used for the first level of recognition, though the nature of the confusion set might vary.

A Novel Hierarchical Classification Scheme for Online Tamil Character Recognition

In this paper we propose a novel three level hierarchical classification scheme for online character recognition for Tamil, a classical Indian language. We make use of the prior knowledge of the writing rules of a Tamil character to build the first level of the classifier for which we outline two methods. The first method utilizes the quantized slope information while the other relies on the trajectory of pen motion for grouping. The number of strokes in the preprocessed character is used for classification at the second level while a k-Nearest Neighbor classifier is employed at the final level. The method that uses the trajectory of the pen motion information is not sensitive to the length of the character and therefore outperforms the method using quantized slope information at the first level of the classifier thereby leading to an increase in the final classification accuracy at the third level from 85% to 96%.

Lexicon-Free, Novel Segmentation of Online Handwritten Indic Words

Research in the field of recognizing unlimited vocabulary, online handwritten Indic words is still in its infancy. Most of the focus so far has been in the area of isolated character recognition. In the context of lexicon-free recognition of words, one of the primary issues to be addressed is that of segmentation. As a preliminary attempt, this paper proposes a novel script-independent, lexicon-free method for segmenting online handwritten words to their constituent symbols. Feedback strategies, inspired from neuroscience studies, are proposed for improving the segmentation. The segmentation strategy has been tested on an exhaustive set of 10000 Tamil words collected from a large number of writers. The results show that better segmentation improves the overall recognition performance of the handwriting system.

DYNAMIC SPACE WARPING OF STROKES FOR RECOGNITION OF ONLINE HANDWRITTEN CHARACTERS

This paper suggests a scheme for classifying online handwritten characters, based on dynamic space warping of strokes within the characters. A method for segmenting components into strokes using velocity profiles is proposed. Each stroke is a simple arbitrary shape and is encoded using three attributes. Correspondence between various strokes is established using Dynamic Space Warping. A distance measure which reliably differentiates between two corresponding simple shapes (strokes) has been formulated thus obtaining a perceptual distance measure between any two characters. Tests indicate an accuracy of over 85% on two different datasets of characters.

Language models for online handwritten Tamil word recognition

N-gram language models and lexicon-based word-recognition are popular methods in the literature to improve recognition accuracies of online and offline handwritten data. However, there are very few works that deal with application of these techniques on online Tamil handwritten data. In this paper, we explore methods of developing symbol-level language models and a lexicon from a large Tamil text corpus and their application to improving symbol and word recognition accuracies. On a test database of around 2000 words, we find that bigram language models improve symbol (3%) and word recognition (8%) accuracies and while lexicon methods offer much greater improvements (30%) in terms of word recognition, there is a large dependency on choosing the right lexicon. For comparison to lexicon and language model based methods, we have also explored re-evaluation techniques which involve the use of expert classifiers to improve symbol and word recognition accuracies.

Online Character Recognition using Regression Techniques

This paper introduces a scheme for classification of online handwritten characters based on polynomial regression of the sampled points of the sub-strokes in a character. The segmentation is done based on the velocity profile of the written character and this requires a smoothening of the velocity profile. We propose a novel scheme for smoothening the velocity profile curve and identification of the critical points to segment the character. We also propose another method for segmentation based on the human eye perception. We then extract two sets of features for recognition of handwritten characters. Each sub-stroke is a simple curve, a part of the character, and is represented by the distance measure of each point from the first point. This forms the first set of feature vector for each character. The second feature vector are the coefficients obtained from the B-splines fitted to the control knots obtained from the segmentation algorithm. The feature vector is fed to the SVM classifier and it indicates an efficiency of 68% using the polynomial regression technique and 74% using the spline fitting method.

Resolving Ambiguities in Confused Online Tamil Characters with Post Processing Algorithms

This paper addresses the problem of resolving ambiguities in frequently confused online Tamil character pairs by employing script specific algorithms as a post classification step. Robust structural cues and temporal information of the preprocessed character are extensively utilized in the design of these algorithms. The methods are quite robust in automatically extracting the discriminative sub-strokes of confused characters for further analysis. Experimental validation on the IWFHR Database indicates error rates of less than 3 % for the confused characters. Thus, these post processing steps have a good potential to improve the performance of online Tamil handwritten character recognition.