MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202621099841 A) filed by Mrs. Kalpana Dhende; Dr. Minakshi More; Dr. Santosh Deshpande; and Dr. Manasi Shirurkar on August 18, 2026, for Deep Learning Signature Verification System.
Inventors include Mrs. Kalpana Dhende; Dr. Minakshi More; Dr. Santosh Deshpande; and Dr. Manasi Shirurkar.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: The present invention relates to a Deep Learning Signature Verification System configured to automatically verify the authenticity of handwritten signatures using image processing and deep learning techniques. The system comprises a signature acquisition module, preprocessing module, deep learning feature extraction module, reference signature database, similarity analysis module, and verification decision module. During enrollment, one or more genuine signature samples of an authorized user are acquired and processed to generate representative numerical feature embeddings, which are securely stored in the reference signature database. During verification, a questioned signature is acquired from a scanned document, digital device, electronic form, or other suitable source and subjected to preprocessing operations including grayscale conversion, noise removal, background removal, cropping, resizing, thresholding, and normalization. The processed signature is subsequently supplied to a trained deep learning model configured to automatically learn distinctive visual and structural characteristics, including stroke patterns, contours, spatial relationships, and overall signature structure. The generated feature representation is compared with a corresponding enrolled reference representation using a similarity or distance measurement technique. A decision-making module evaluates the resulting similarity value against a predetermined or dynamically determined verification threshold and generates a genuine, forged, rejected, or unverified result. A confidence value may additionally be generated to indicate the strength of the verification decision. The system reduces dependence on manual signature inspection and handcrafted feature selection while accommodating natural variations in genuine signatures. The invention is applicable to banking, financial services, insurance, government documentation, educational institutions, legal records, healthcare, corporate approvals, and electronic document authentication systems.
Disclaimer: Curated by HT Syndication.
