USPTO Patent US12581463B1: Spectral Detection of Radio Events

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 17th, 2026

Detected March 23rd, 2026

Summary

The USPTO has granted patent US12581463B1 to Virginia Tech Intellectual Properties, Inc. for methods and systems related to spectral detection and localization of radio frequency (RF) signals using machine-learned convolutional neural features. The patent covers techniques for classifying RF signals based on learned features and comparing them to known signal classes.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has issued patent US12581463B1, titled 'Spectral detection of radio events,' to Virginia Tech Intellectual Properties, Inc. The patent details methods and apparatus for training and deploying machine-learned classification of radio frequency (RF) signals. Key aspects include obtaining RF spectrum data, segmenting it into samples, and using a machine-learning network to compare information within each sample to labeled signal classes, ultimately identifying and matching the information to a specific class.

This patent grant represents a new intellectual property asset in the field of AI-driven signal processing. While it does not impose new regulatory obligations on companies, it signifies innovation in AI applications for radio event detection. Companies operating in telecommunications, defense, and technology sectors that utilize RF spectrum analysis or AI for signal classification should be aware of this granted patent, as it may impact their freedom to operate or potential licensing requirements.

Source document (simplified)

← USPTO Patent Grants

Spectral detection and localization of radio events with learned convolutional neural features

Grant US12581463B1 Kind: B1 Mar 17, 2026

Assignee

Virginia Tech Intellectual Properties, Inc.

Inventors

Timothy James O'Shea, Tamoghna Roy

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training and deploying machine-learned classification of radio frequency (RF) signals. One of the methods includes obtaining input data corresponding to the RF spectrum; segmenting the input data into one or more samples; and for each sample of the one or more samples: obtaining information included in the sample, comparing the information to one or more labeled signal classes that are known to the machine-learning network, using results of the comparison, determining whether the information corresponds to the one or more labeled signal classes, and in response, matching, using an identification policy of a plurality of policies available to the machine-learning network, the information to a class of the one or more labeled signal classes, and providing an output that identifies an information signal corresponding to the class matching the information obtained from the sample.