AI Drone Detection Through Frequency Spectrum Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for detecting drones are inaccurate and unreliable due to encrypted signals, making it difficult to determine their presence and implement effective defense mechanisms, especially in restricted airspace.

Innovation Solution

A method using AI-based machine learning models, such as support vector machines (SVMs), to analyze drone signals without decoding, by recognizing characteristic frequency spectra and patterns in radio signals, including DroneID, video, and remote control signals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If radar is used to detect drones through encrypted signals, then detection coverage is achieved, but measurement precision deteriorates

Engineering Contradiction:
Improvedrone detection reliabilityVSAvoiddrone detection precision
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent replaces radar-based physical detection with an information-processing approach using machine learning. Instead of relying on electromagnetic wave reflection (radar), the system processes radio signal characteristics through trained AI models to identify drones, thereby improving measurement precision while maintaining reliable detection

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

Solution Approach 2:

The patent transforms the detection approach by changing from physical parameter measurement (radar range, velocity) to signal parameter analysis (frequency spectrum, temporal patterns). By analyzing multiple signal parameters through machine learning, the system achieves both reliable detection and precise identification of drone presence

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If signal decoding is performed to identify drone manufacturer, then detection precision is improved, but loss of time increases

Engineering Contradiction:
Improvedrone identification accuracyVSAvoidsignal processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training machine learning models with extensive drone signal data before deployment. The models learn to recognize manufacturer-specific signal patterns in advance, enabling rapid identification without real-time decoding. This pre-processing eliminates time loss during actual detection while maintaining high precision

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating trained machine learning models that replicate the complex decoding and analysis processes. Instead of performing full signal decoding in real-time, the system uses the trained model copies to rapidly classify signals based on previously learned patterns, significantly reducing processing time while maintaining identification accuracy

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4636614A1Drone detection method, method for training ai model, computer program and sensing device for performing the detection method
Publication Date: 2025.10.22 BUNDESDRUCKEREI GMBH
  • EP4636614A1 patent drawingFigure 1
  • EP4636614A1 patent drawingFigure 2~3
  • EP4636614A1 patent drawingFigure 4

AI summary

The invention relates to an AI-based method for detecting a drone (100), comprising the steps of: - detecting a plurality of wirelessly transmitted radio signals (200) over the air using at least one detection antenna of a detection device; - transforming the radio signals (200) into a frequency spectrum (202); - analyzing the frequency spectrum (202) using an image-based AI model (400) trained on drone signals (300, 302, 304); and - determining the presence of at least one drone (100) within the range of the detection device if the image-based AI model (400) detects a match or if the image-based AI model (400) detects at least one match up to a predetermined threshold. The invention also relates to a method for training the AI ​​model (400), a computer program and a detection device for carrying out the detection method.