MUMBAI, India, Oct. 19 -- Intellectual Property India has published a patent application (202611096510 A) filed by Noida Institute Of Engineering And Technology Niet on August 10, 2026, for Edge-Deployable Neural Channel Emulation System With Domain-Decomposed Representation Learning.
Inventors include Rahul Agarwal; and Ravi Anand.
The application for the patent was published on October 16, 2026, under issue no. 42/2026.
Abstract: An edge-deployable neural channel emulation system (100) and method (200) for synthesizing standards-compliant wireless channel realizations on resource- constrained edge hardware. A channel data acquisition module (110) normalizes field measurements from base station antenna receivers into canonical tensors. A domain-decomposed representation engine (120) comprising temporal (122), spatial (124), and frequency (126) representation units independently extracts features along each physical dimension; a hierarchical fusion unit (128) couples these features through pairwise bilinear attention. An edge inference optimization module (130) applies knowledge distillation (132), quantization-aware training (134), and hardware-adaptive partitioning (136) for deployment on edge devices equipped with neural processing units. A standards-ready channel emulator (140) with a generative synthesis unit (142) and a standards compliance mapping unit (144) produces realizations formatted for standardized protocol stacks. A closed-loop calibration module (150) refines accuracy against field measurements. A federated coordinator (160) enables privacy-preserving multi-site learning. [Figure 1]
Disclaimer: Curated by HT Syndication.
