Title :
OLDA based online handwriting recogntion
Author :
Prasad, M. Mahadeva
Author_Institution :
Dept. of Studies in Electron., Univ. of Mysore, Hassan, India
Abstract :
The paper presents the work carried out in recognizing online handwritten data using OLDA. A comparative study on the performance of OLDA and RLDA in terms of recognition accuracy and recognition speed is carried out. Online handwritten Kannada basic character data is used for the experiments. Writer independent experiments are carried with 3750 samples for training and 1550 samples for testing. With estimate feature and nearest neighbor as a classifier, an average maximum recognition accuracy of 88.7% and 88.5% has been achieved with OLDA and RLDA respectively. While OLDA has achieved the best recognition accuracy with only 20 eigen vectors, RLDA has taken 25 eigen vectors. The experiments reveal that the performance of OLDA is better than that of RLDA in terms of recognition accuracy, computation cost and also the memory requirements.
Keywords :
handwritten character recognition; image classification; natural language processing; statistical analysis; OLDA based online handwriting recognition; RLDA; average maximum recognition accuracy; memory requirements; nearest neighbor classifier; online handwritten Kannada basic character data; online handwritten data recognition; writer independent experiments; Accuracy; Algorithm design and analysis; Classification algorithms; Feature extraction; Support vector machine classification; Testing; Training; Kannada character recognition; online handwriting recognition; orthogonal linear discriminant analysis; regularized linear discriminant analysis;
Conference_Titel :
Contemporary Computing and Informatics (IC3I), 2014 International Conference on
Conference_Location :
Mysore
DOI :
10.1109/IC3I.2014.7019801
