MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073861 A) filed by K Sakthiyavathi; Dr. K. Saruladha; and S. V Jeyasri on June 15, 2026, for Reynolds Number Based Model For Lie Detection Using Computer Vision.
Inventors include K Sakthiyavathi; Dr. K. Saruladha; and S. V Jeyasri.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The invention provides a computer vision-based system for detecting deception from facial micro~expressions in real-life trial videos. The method starts by extracting frames from input videos and identifYing the onset and apex frames, which correspond to the start and peak of facial expressions. The apex frame is determined using Local Binary Pattern (LBP)-based feature difference analysis. The Optical flow estimation technique RAFT captures the motion between . the onset and apex frames to detect subtle pixel motion patterns in facial expressions. A fluid dynamics-inspired Reynolds feature map is computed by integrating motion velocity, spatial characteristics, and gradient information. The obtained Reynolds feature map is fed into the Block Division Convolutional Neural Network (BDCNN) for facial feature extraction. The extracted features are combined, and 1 x 1 convolutions are applied for dimension reduction, followed by fully connected layers. The system outputs a decision indicating whether the behaviour is truthful or deceptive, improving accuracy in detecting subtle facial expressions.
Disclaimer: Curated by HT Syndication.
