MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641107187 A) filed by R Aravind; Dr. M. Sughasiny; and Mr. M Mily Maarshiyan on September 07, 2026, for System And Method For Real-Time Tamil Speech-To-Tamil Fingerspelling Sign Translation Using Confidence-Adaptive Linguistic Mapping.
Inventors include R Aravind; Dr. M. Sughasiny; and Mr. M Mily Maarshiyan.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: [050] A system and method are disclosed for substantially real-time translation of Tamil speech into visual Tamil sign output. The system receives live or recorded Tamil audio, performs speech conditioning, generates Tamil text with confidence information, and determines stable textual units from evolving speech-recognition hypotheses. The stable text is canonically normalized to preserve Tamil grapheme or orthographic character boundaries. A confidence-adaptive mapping engine selects a validated lexical sign when availability and confidence criteria are satisfied and, when a lexical sign is unavailable or uncertain, automatically decomposes the textual unit into Tamil grapheme or character units and retrieves corresponding fingerspelling assets. A sequence composer orders the assets using transition metadata, while a latency controller bounds delay and prevents repeated rendering of revised partial hypotheses. Unknown lexical content is thereby preserved through deterministic character-level fingerspelling instead of being silently omitted. The architecture is implementable using static images, video assets, skeletal poses or a sign-rendering avatar and is suitable for assistive communication in education, healthcare, workplaces and public services. Accompanied Drawing [FIGS. 1-2].
Disclaimer: Curated by HT Syndication.
