Texture-Based Author Identification: A Novel Approach Using Gabor Filtering and Grey Scale Co-occurrence Matrices

Abstract

This research introduces a unique method for doing authorship verification on handwritten papers using texture analysis. Our technique is content-independent for any given script, as opposed to the more common approaches that rely on segmentation and the written text itself. For author identification, we use weighted Euclidean distance and K-nearest neighbour classifiers, while for texture feature extraction, we make use of 2-D Gabor filtering and grey scale co-occurrence matrices. The findings of our research on samples of handwriting in other languages are encouraging, demonstrating the promise of this texture-based method to biometric identification.