Writer Identification of Bangla Handwritings by Radon Transform Projection Profile

Writer identification is the task of determining the person whose handwritten sample is available in a set of writings, collected from multitude of writers. This has useful applications in many areas, notably in forensic analysis. The task of writer identification is quite difficult due to minimal variations found in different handwritten samples from same person/writer. Several identification algorithms have been proposed so far which are mostly for non-Indic writings. This paper presents a new approach for extracting two different sets of components (essentially fragments of characters), namely fragment set-A and fragment set-B. Features are extracted from each element of these two sets to identify the writing style of a particular person. The features are computed based on Radon transform projection profile. The proposed approach uses lesser amount of information from the handwritten samples, thus saving computation time as well as memory requirement. The condition to determine that the writer is unknown (i.e., there is no handwritten sample from that writer in reference base) is also proposed. The approach is tested on a collected dataset of Bangla writings and the experimental results are encouraging.