MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641049935 A) filed by Prasad V. Potlurl Siddhartha Institute Of Technology on April 20, 2026, for Ai Powered Assistive Mobility Tool For Visually Impaired.
Inventors include Dr. S Parvathi Vallabhaneni; K Santhi Priya; B Nikhitha; G Mounika; and K L Janardhan.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: A software platform is disclosed that addresses the navigational needs of visually impaired individuals through the integration of real-time object detection, monocular distance estimation, horizontal position mapping, and closed-loop voice interaction, with all operations performed locally on the user's device without any dependency on external networks. A YOLOv8 Nano detection model - achieving 100% mean average precision across eighty COCO object categories at 1100 frames per second - processes frames captured from a device camera. Each detected object is assigned a proximity band (close, nearby, or far away) derived from the thin-lens distance formula using canonical object dimensions, and a horizontal zone (far left, slightly left, center, slightly right, or far right) derived from the proportional position of the bounding box center. A natural-language audio script, constructed from these spatial attributes and prioritized by proximity, is delivered through an offline text-to- speech engine. A voice interaction state machine governs a continuous listen-speak cycle by enforcing strict temporal separation between synthesis output and microphone activation. The cycle is paused automatically when the detection camera is active, preventing spoken detection announcements from contaminating the microphone input. Color-coded bounding box overlays rendered on an HTML5 Canvas element provide simultaneous visual feedback for partially sighted users or observers. The platform comprises a web application built on Node.js, Python Flask, MongoDB, and JavaScript, as well as a React Native mobile application that performs all AI inference on-device through TensorFlow.js, stores data in SQLite, and requires no server connection during operation. Both platforms share a unified design philosophy of complete offline functionality, authenticated user sessions, and detection history logging, making this invention a comprehensive and deployable solution for real-world assistive mobility
Disclaimer: Curated by HT Syndication.
