Writer Identification using Handwriting Samples
Spurthi S. Bhat, Vaishnavi Bhokare, Rutuja Bhirud, Pushkar S. Joglekar · 2022
The writer identification task has received a lot of interests due to its variety of applications in forensic analysis and security systems. However, research studies have focused on offline handwritten document analysis and mostly in English handwritten scripts. The research in the field of Indian Languages is very limited because of lack of suitable databases. The proposed methodology for writer identification is based on Bangla handwriting script and consists of two phases: Feature extraction using Local Binary Pattern and Machine Learning algorithms using Support Vector Machine and K-Nearest Neighbor classifiers. The average accuracy achieved through this method is 93.34% through Support Vector Machine and 90% through K-Nearest Neighbor algorithm. Different experiments have been carried out in terms of number of writers, types of scripts etc. to analyze the impact of various factors in the accuracy of the proposed algorithm.