MUMBAI, India, July 5 -- Intellectual Property India has published a patent application (202311086534 A) filed by Samsung Electronics Co. Ltd., Gyeonggi, Republic of Korea, on Dec. 18, 2023, for 'method and system for processing disfluent speech and estimating time of completion.'
Inventor(s) include Sachdeva, Kapil; Chowdhury, Sudip; Jangid, Sachin; and Balyan, Shavik.
The application for the patent was published on July 4, under issue no. 27/2025.
According to the abstract released by the Intellectual Property India: "Disclosed herein is a method (1000) for generating speech endpoint for a disfluent type of speech. The method includes detecting an utterance of a user, the utterance including the input audio stream having plurality of audio chunks. Subsequently identifying a speech pattern of the user based on the acoustic features of each of the plurality of audio chunks and further identifying spectral and temporal characteristics of audio chunks. Subsequently, the method (1000) includes generating a contextual embedding of a previous audio chunk preceding a current audio chunk. Further, the method includes determining a location of one or more disfluent audio chunks. The method further includes predicting an expected endpoint of each of the disfluent audio chunks. Consequently, updating the EPD time of the VA based on the determined type of disfluent audio chunks and the predicted endpoint of each of the disfluent audio chunks."
Disclaimer: Curated by HT Syndication.
