MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611072772 A) filed by Noida Institute Of Engineering And Technology Niet on June 11, 2026, for Federated Neural Signal Processing Apparatus With Hardware-Isolated Normalization And Dual- Space Adversarial Defense.
Inventors include Dr. Rahul Kumar Sharma; and Vivek Kumar Sharma.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: A federated neural signal processing apparatus (100) for privacy-preserving brain-computer interface classification comprises a central aggregation server (110) and a plurality of client neural signal processing nodes (120) connected via a parameter distribution bus (114). Each client node (120) includes an electroencephalogram acquisition front-end (122), a local normalization engine (124) with batch statistics registers (136), an adversarial perturbation generator (126) for input-space hardening, an adversarial weight perturbation module (128) for parameter-space robustness, and a local model parameter memory (130). A hardware write-lock latch (132) governed by a model training controller (146) maintains a locked state during training and unlocks only after parameter restoration completes, preventing a parameter upload controller (140) from accessing intermediate states or batch statistics. A bus firewall (134) with a latched master-ID filter permits upload of only non- normalization parameters, enabling cross-subject generalization, adversarial robustness, and data privacy through hardware-enforced isolation.
Disclaimer: Curated by HT Syndication.
