MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641075787 A) filed by Vardhaman College Of Engineering on June 18, 2026, for Method For Blind Source Separation In High-Interference Signal Environments.
Inventors include Ch. Sreenivasa Rao; Dr. V. Purna Chandra Reddy; K. Sripal Reddy; K. Ashwini; Sunkaraboina Sreenu; and A. Vijaya Lakshmi.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: Method for Blind Source Separation in High-Interference Signal Environments is the proposed invention. The proposed invention discloses a method for blind source separation in high-interference signal environments utilising transformer-based deep learning, self-supervised representation learning, and probabilistic latent-space modelling techniques. The proposed system collects the multidimensional mixed signals from the complex sensing and communication environment, and performs intelligent pre-processing, such as spectral decomposition, denoising, synchronisation and feature extraction. A transformer-based signal analysis engine with multi- head attention mechanisms learns temporal and spectral dependencies between mixed signal components to identify hidden source structures. Self-supervised learning and variational probabilistic decomposition models allow the separation and reconstruction of independent source signals, without needing any prior knowledge on source characteristics or mixing conditions. Graph neural networks and reinforcement learning mechanisms further optimise the separation accuracy, adaptive filtering and interference suppression under dynamic operating conditions. The invention improves signal recovery performance, robustness, scalability and real-time adaptability for wireless communication, biomedical diagnostics, radar imaging, sonar systems and intelligent signal processing applications.
Disclaimer: Curated by HT Syndication.
