The paper discusses the performance of the online handwritten recognition module designed for Kannada numerals. In the proposed system, there is an improvement in the recognition accuracy of the recognition module when compared to the previous ones. The difference feature has given the improved performance of the recognition module. The difference feature vector is formed by computing the difference of consecutive x- and y- coordinate values of online handwriting. The writer independent experiments are carried out by dividing the collected online handwritten Kannada numeral data into a disjoint set of training and testing data sets. Out of 1400 numeral data samples, 700 numeral data samples are used for training the recognition modules and the remaining 700 numeral data samples are used for testing. The collected online handwritten data are subjected to preprocessing and feature extraction. The difference feature is extracted from the preprocessed data. The extracted features are mapped to lower-dimensional feature space by using OLDA. The subspace features are used to train and test the recognition module. Classification of the test data is carried out by using the nearest neighbor classifier. The average recognition accuracy of 99.3% is achieved.

Online Handwriting, Kannada Numerals; Handwriting Recognition Module; Difference Feature; Nearest Neighbor Classifier

[1] R O Duda, P E Hart, and D Stork, Pattern Classification, Second Edition, John Wiley and Sons (Asia) Pvt. Ltd., 2006.
[2] A.L. Blum, P. Langley, “Selection of relevant features and examples in machine learning,” Artificial Intelligence, Vol. 97, pp. 245–27, 1997.
[3] Claus Bahlmann, “Directional features in online handwriting recogntion,” Pattern Recognition, Vol. 39, pp. 115-125, 2006.
[4] S. Jaeger, S Manke, J Reichert, and Waibel, “Online handwriting recognition: the NPen++ recognizer,” Int. Jl. Of Document Analysis and Recogntion, Vol. 3, pp.169-80, 2001.
[5] Basabi Chakraborty, Goutam Chakraborty, “A new feature extraction technique for on-line recognition of handwritten alphanumeric characters,” Information Sciences, Vol. 148, pp. 55–70, 2002.
[6] Michael Blumenstein, XinYu Liu, Brijesh Verma, “An investigation of the modified direction feature for cursive character recognition,” Pattern Recognition, Vol. 40, pp. 376-388, 2007.
[7] Wei Zeng, XiangXu Meng, ChengLei Yang, Lei Huang, “Feature extraction for online handwritten characters using Delaunay triangulation,” Computers & Graphics, Vol. 30, pp. 779–786, 2006.
[8] Fengxi Song, Shuhai Liu, Jingyu Yang, “Orthogonilzed Fisher disriminant”, Pattern Recognition, Vol. 38, Issue 2, pp. 311-313, 2005.
[9] M. Mahadeva Prasad, M. Sukumar, A. G. Ramakrishnan, “Divide and conquer technique in online handwritten Kannada character recognition,” In the Proc. of Int. Workshop on Multilingual Optical Character Recognition Systems, 2009, Barcelona, Spain.
[10] M. Mahadeva Prasad, M. Sukumar and A.G. Ramakrishnan, “Orthogonal LDA in PCA transformed subspace,” In the Proc. of 12th Conf. on Frontiers in Handwriting Recognition, pp. 172-175, 2010.
[11] M. Mahadeva Prasad, M Sukumar, “HMM based Two-Stage Classification Scheme to Improve Online Handwritten Kannada Numeral Recognition,” Int. Jl. of Computer Science and Technology, Vol. 3, Issue 2, pp. 897-902, 2012.
[12] M. Mahadeva Prasad, “Writing Direction Feature based Online Handwritten Kannada Numeral Recognition,” Int. Jl. of Computational Engineering Research, Vol. 9, No. 2, pp. 65-68, 2019.