MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096877 A) filed by Vellore Institute Of Technology on August 11, 2026, for Multimodal Sentiment Analysis System With Fuzzy Trust Control.
Inventors include Dr. Prabhavathy P; Dhruv Asodaria; and Ansh Jerath.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT MULTIMODAL SENTIMENT ANALYSIS SYSTEM WITH FUZZY TRUST CONTROL A multimodal sentiment analysis system (100) comprises an atmospheric meta-mining module (110) configured to receive raw sensor data streams including an audio data stream and video data stream, compute a spectral flatness measure from the audio data stream representing a ratio of a geometric mean to an arithmetic mean of a power spectrum, and extract a visual confidence score from the video data stream based on face detection probability. A fuzzy trust controller (120) receives the spectral flatness measure and visual confidence score, applies linguistic rules, and generates normalized scalar trust weights including audio, video, and text trust weights summing to 1.0. Independent modality encoders (130) process the raw sensor data streams in parallel isolation to generate modality-specific sentiment predictions. A weighted ensemble fusion module (140) combines the modality-specific sentiment predictions using the trust weights to generate a final sentiment prediction. (FIG. 1)
Disclaimer: Curated by HT Syndication.
