MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202611100411 A) filed by Chitkara University on August 19, 2026, for An Offline Bidirectional Sign–speech Translation System With Confidence-Controlled Edge Processing.
Inventors include Ojas Kashyap; Ashita Singla; Taniya; Shorya; Dr. Rubina Dutta; and Dr. Parminder Kaur.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention an offline bidirectional assistive communication system (100) comprises an embedded edge processing unit (104), camera (106), microphone (112), manual input device (114), display (150), and speaker (140). In a sign-to-speech mode, a frame-acquisition controller (118) captures hand gestures, an image- preprocessing engine (120) generates normalized frames, and a hand-landmark extractor (122) and pixel-feature extractor (124) provide complementary features to a feature-fusion buffer (126). A temporal gesture stabilizer (128), gesture classifier (130), and confidence evaluator (132) suppress unstable predictions before a text- composition buffer (136) drives an offline text-to-speech engine (138). In a speech-to-sign mode, an offline speech-to-text engine (142), token normalizer (144), sign- mapping engine (146), and visual sign renderer (148) generate sign representations on the display (150). An ambient-light sensor (110), controllable illumination source (108), processor-load monitor (176), latency monitor (178), and model-selection controller (180) maintain capture quality and real-time performance while a privacy switch (168) permits network-isolated operation during portable use. (Accompanied Figure No. 1 to 4)
Disclaimer: Curated by HT Syndication.
