A fuzzy Approach for the Automatic Off‑line Arabic Signature Verification Problem

Offline Arabic signature verification (OSV) is a challenging biometric task due to the high stylistic variability of Arabic handwriting and the presence of skilled forgeries. This work proposes a hybrid verification system that integrates geometric feature extraction with a Takagi-Sugeno fuzzy inference model. After preprocessing, the system extracts the skeleton of each signature and detects key control points using the Shi-Tomasi algorithm. Four discriminative local geometric features-distance deviation, angular deviation, proportional distance ratio, and centroid deviation-are computed between matched control‑point pairs of the reference and test signatures. These features capture subtle structural inconsistencies introduced by genuine handwriting variation or forgery attempts. A fuzzy inference system with sixteen rules maps these features into a similarity score, and a writer‑dependent thresholding mechanism determines acceptance or rejection. Experiments conducted on a dataset of 50 Arabic writers demonstrate that the proposed method achieves competitive accuracy, reduces false acceptance and rejection rates, and provides an interpretable framework suitable for forensic and banking applications.