A hybrid approach for off-line cursive script recognition

• Main Author: Monreal, Justin T.
• Format: text
• Language: English
• Published: Animo Repository 1998
• Subjects: Writing; Image processing; Character sets (Data processing); Neural networks (Computer science); Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/2495; https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=9333&context=etd_masteral
• Institution: De La Salle University

Cursive script recognition is commonly based on finding letters within a word and recognizing them separately. The segmentation process is ambiguous and difficult. This paper presents a hybrid method which combines individual recognizers: segmentation-based and word-based, to cope with difficulties in recognizing cursive script. Words are first segmented into smaller subimages. A neural network is used to identify possible letters among the group. Letter information is combined with word shape information to get word identity. Recognition results of individual and hybrid recognizers are presented. The hybrid recognizer is found to perform better than individual recognizers.