Online Handwritten Telugu Stroke Recognition using Direction Operator for Feature Extraction
1Srilakshmi Inuganti, Dr. R. Rajeshwara Rao
Important phase of Character Recognition is Feature extraction. In this paper a feature extraction technique using 8-point interconnection method named as Direction operator is proposed. The feature obtained by using direction operator represents structural, global and local characteristics of a stroke. This feature represents (x,y) coordinate of each point on the stroke with 8-bit vector, where each vector gives interconnection of this point with other points in the stroke in 8-directions.A two stage classifier known as Hybrid classifier is also proposed, in which direction operator feature is used in preclassifier and preprocessed(x,y) coordinates are used in postclassifier. We have verified this feature with HP-Lab data available in UNIPEN format. Experimental results proved that direction operator feature improves recognition accuracy over the chosen dataset.
Direction operator, Hybrid Approach, Stroke Recognition, Online