Intellectual Property India Publishes Patent Application for Multimodal Deep Learning System For Emotional Recognition From Speech And Facial Expressions

MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112451 A) filed by Dr. R. Caroline Kalaiselvi; Dr. G. Kiruthiga; Dr. G. Preetha; Dr. K. Raviya; Dr. I. Preethi; and Dr. S. Nirmala Devi on September 18, 2026, for Multimodal Deep Learning System For Emotional Recognition From Speech And Facial Expressions.

Inventors include Dr. R. Caroline Kalaiselvi; Dr. G. Kiruthiga; Dr. G. Preetha; Dr. K. Raviya; Dr. I. Preethi; and Dr. S. Nirmala Devi.

The application for the patent was published on September 25, 2026, under issue no. 39/2026.

Abstract: The present invention discloses a multimodal deep learning system for recognizing human emotions from speech and facial expressions. The system receives synchronized or corresponding audio and facial-video inputs and independently preprocesses the respective modalities. A speech deep-learning network generates speech-emotion representations, while a facial deep-learning network generates spatial-temporal facial-expression representations. A temporal synchronization module aligns the representations, and an adaptive multimodal fusion module combines them according to learned relevance and modality reliability. A classification module processes the fused representation to generate an emotional state and confidence value. The system further supports noisy, incomplete, and variable-quality audiovisual inputs for robust emotion recognition.

Disclaimer: Curated by HT Syndication.