AI Spectrum Categorization via Image Conversion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional spectrum analysis methods rely on human operators to manually inspect numerical data, which is time-consuming, error-prone, and may miss signals due to limitations in human perception and processing.

Innovation Solution

The method involves converting processed spectrum data into digital image files, allowing for the use of convolutional neural networks (CNNs) for AI-based spectrum categorization, thereby improving efficiency, accuracy, and reliability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional manual inspection methods are used for spectrum analysis, then human operators can detect signals, but the process is time-consuming and error-prone

Engineering Contradiction:
Improvedetection accuracyVSAvoidanalysis time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces the mechanical human inspection process with an automated computer-based system that converts spectrum data into image format and applies convolutional neural networks for automatic classification, eliminating manual intervention while improving both speed and accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary image conversion step that transforms numerical spectrum data into visual image format, enabling the use of powerful image recognition algorithms (CNNs) that can rapidly and accurately classify signals without direct human intervention

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If numerical data processing methods are used for spectrum categorization, then the process can be automated, but the accuracy and efficiency are limited compared to image-based AI analysis

Engineering Contradiction:
Improvecategorization efficiencyVSAvoidcategorization accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent changes the data representation parameter from numerical values to image format, transforming the spectrum data into a visual representation that can be processed by CNNs, thereby simultaneously improving both processing efficiency and categorization accuracy

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent adds a visual dimension to the spectrum data by converting it into image format with spatial and frequency dimensions, enabling the application of two-dimensional convolutional operations that are highly efficient for pattern recognition and significantly improve categorization performance

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP4571580A1Method and device for artificial intelligence-based spectrum categorization
Publication Date: 2025.06.18 ROHDE & SCHWARZ GMBH & CO KG
  • EP4571580A1 patent drawingFigure 1
  • EP4571580A1 patent drawingFigure 2
  • EP4571580A1 patent drawingFigure 3

AI summary

A device for artificial intelligence-based spectrum categorization is provided. The device allows for identifying different signal types present within the spectrum and automatically categorizing them using image based artificial intelligence algorithms. The device is configured to receive a radio signal comprising an IQ sample, calculate a spectrogram of the IQ sample, convert the spectrogram into an image, and provide the image to a pre-trained convolutional neural network, CNN, stored in the device. The CNN is configured to determine, whether the radio signal corresponding to the image belongs to one of the following categories: noise floor, target signal, interference signal, target signal with interference signal.