Since digitization is yet to be adopted globally, handwritten documents are still in use in many places. Handwritten documents are prone to get forged thanks to acts like the versatility of tampering which are very frequent among skilful fraudsters. Our research work focuses on one of the major problems to detect whether a document is treated as false or not based on an analysis of the handwriting of the content writers. Mostly, legal documents are scripted authentically by a single person. If the content is a combination of more than one person, then it will be treated like a forged document. The proposed work is formulated as a binary classification problem. Various contour related sliding window based features are extracted from word images of the corresponding handwritten document. The same writer with different handwriting styles are also considered here as well. Bagging meta-classifier is trained for classification of the extracted features. The accuracy of this proposed work is 89.64% on IAM dataset is quite sound. We have also tested our method on IDRBT check image dataset. However, since there is a lack of direct implementation on this particular problem we could not make a comparative analysis of the proposed method.
展开▼
