MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109751 A) filed by Bharathi N; Dhivyadharshini N S; Jeevika S; Vivekanandha College Of Engineering For Women; Ms. S. Sinduja; Ms. R. S. Mathubala; and Ms. S. Fowjiya on September 12, 2026, for A System And Method For Sign Language Recognition And Communication Using Cnn-Lstm And Text- To-Speech.
Inventors include Bharathi N; Dhivyadharshini N S; Jeevika S; Vivekanandha College Of Engineering For Women; Ms. S. Sinduja; Ms. R. S. Mathubala; and Ms. S. Fowjiya.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: The invention relates to artificial intelligence based assistive communication, particularly real-time sign language recognition using computer vision and deep learning. A webcam captures hand gestures as video frames, MediaPipe extracts 21 three-dimensional hand landmark coordinates, and a preprocessing stage normalizes and denoises the landmark data to reduce variations due to hand size, pose, lighting, and camera angle. A CNN extracts spatial features of hand shape, finger orientation, joint position, and gesture pattern, while an LSTM learns temporal relationships across successive frames to recognize static as well as continuous gestures. The classified gesture is converted into meaningful text and further into audible speech through Text-to-Speech, with optional real-time transmission to a mobile application. The invention provides a contactless, low-cost, integrated communication platform without sensor gloves or specialized hardware, for use in healthcare, education, workplaces, public services, and home environments. Figure 2 best characterises the invention.
Disclaimer: Curated by HT Syndication.
