Writer identification is a quite interesting researchproblem in the field of writing recognition due to an ambiguouswritten styles of different writers. This paper proposes amodel that hybridises Convolutional Neural Network (CNN)and Multiclass-Support Vector Machine (MSVM) for getting abetter accuracy in writer identification using English/Arabichandwriting samples. Deep identifying writer takes local handwritten image as input and CNN used for feature extractionthen classified using MSVM classifier based on the extractedfeatures from the CNN layers. The used CNN architecturewas applied with multiple kernel sizes and each time thecorresponding processing time and the identification accuracywas measured. The proposed system was applied over twopublicly databases Khatt as an Arabic database and IAM asan English database and able to achieve an accuracy of around99.8% for a set of 206 writers. The performance of the proposedsystem was compared with other existing writer identificationsystems.
展开▼
