MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202621075799 A) filed by D. Y. Patil International University on June 18, 2026, for Dysarthric Speech Conversion Apparatus With Real-Time Intelligibility Estimation And Adaptive Acoustic Feature Extraction.
Inventor includes Laxmi Priya Sahu.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: ABSTRACT The present disclosure discloses a dysarthric speech conversion apparatus (100) comprising a microphone (102) for capturing dysarthric speech, a pre- processing module (104) for producing isolated speech segments, and a dynamic precision acoustic feature extractor (106) configured to compute time-frequency representations at a plurality of selectable numerical precision levels. An intelligibility estimation module (108) estimates a real-time intelligibility metric of the speech signal, and a precision control mechanism (110) dynamically selects the numerical precision level based on the intelligibility metric. A user-adaptive neural speech conversion model (112) receives the dynamically precision-controlled acoustic features and generates intelligible speech output or textual output conditioned on a user- specific dysarthria embedding. An on-device adaptation module (120) updates the embedding and model parameters using supervised correction examples without requiring off-device computation. The apparatus (100) delivers converted outputs through an output interface (114) to a communication partner. FIG. 1
Disclaimer: Curated by HT Syndication.
