Handwritten script identification using possibilistic approach for cluster analysis

Abstract

One very important field of application of pattern recognition is character recognition. But, until now, almost all character-recognition systems make an implicit assumption that the script and/or the language of the document to be processed is known. This may not be true in an intenational environment. So, script identification plays an importantrole in automatic processing of document images. In this paper, we present a new scdpt identification algorithm that can identify the same from the characters in a handwritten document. The proposed scheme is based on training viaclustering and classification using possibilistic membership grades of a character to all the script classes. This enablesthe recognizer to work with ambiguous and uncertain cases. The recognizer may reject a pattern when it can not make any clear decision ensuring high reliability. Simulation results confirm the effectiveness of the. proposed script identification technique