STO-Activities: (no title)
Activity title:
Sparse Representations and Machine Learning for Radio Frequency Signal Processing
Activity Reference:
SET-350
Panel:
SET
Security Classification:
NATO UNCLASSIFIED
Status:
Proposed
Activity type:
RTG
Start date:
2025-02-02T00:00:00Z
Actual End date:
2028-02-02T00:00:00Z
Keywords:
Classification, Compressed Sensing, Deep Learning, Detection, Machine Learning, Radar, RF, SET, Signal Processing, Sparse Representations
Background:
Radio frequency (RF) signal processing is an extensive and imperative engineering field that affects performance of NATO systems that enable multi-domain operations such as satellite communications, space situational awareness, missile defense, target detection, tracking and classification, amongst others. There is a growing need to field novel RF signal processing techniques that rapidly and reliably process large amounts of incoming data for multiple missions to support the NATO Warfighting Capstone Concept’s Warfighting Development Imperatives (WDI) of Cognitive superiority in RF sensing systems, integrating multi-domain defense, and layered resilience.
In recent years Machine Learning (ML) has achieved tremendous success in many commercial applications such as automatic face recognition, speech recognition, natural language processing, autonomous vehicles, and robotics. An important element in the success of ML for these applications is the availability of large, labeled databases which are used in training. However, in many military applications, large sets of labeled training data are unavailable. This is especially true in RF sensing applications. As a result, sparse representations of RF sensing data must be considered.
Thus, in this activity we will explore convergent technologies that combine sparse representations and machine learning in a synergistic and novel manner to create a disruptive effect. We go beyond classic radar applications to broadly consider the field of RF signal processing. We systematically compare classical techniques with modern machine learning approaches in controlled environments while exploring topics such as the detection of small, closely spaced targets within range-doppler signatures that contain non-Gaussian sea clutter, the categorization of micro-Doppler signatures for UAV classification, waveform modulation parameter estimation using machine learning and the application of techniques to multi-function RF systems.
Objectives:
The main objectives of the RTG are to:
• Explore the development of novel machine learning-based RF signal processing techniques for Space, Air, Ground and Maritime applications in support of Multi-Domain Operations
• Compare classic sparse signal analysis techniques to modern machine learning approaches in controlled experiments.
• Expand existing radar analyses to more generic RF applications.
• Develop labeled datasets and novel algorithms for micro-Doppler signature classification, target detection in non-Gaussian clutter, and modulated waveform characterization.
The results of the RTG will be described in a technical report to be delivered before the end of the RTG.
Topics:
Themes of convergence in sparse representations and ML that will be addressed in this NATO group include:
• Physics informed ML algorithms, complex-valued neural networks and federated learning for RF signal processing.
• Novel/improved algorithms for detection of small, closely spaced targets within range-doppler signatures that contain non-Gaussian sea clutter.
• Categorization of micro-Doppler signatures for UAV classification.
• Waveform modulation parameter estimation using machine learning.
• Application of techniques to multi-function RF systems.
Contact:
Open2Partners:
Title:
Created at 23/04/2024 19:01 by System Account
Last modified at 16/05/2024 16:00 by System Account
Go back to list
Home(NATO STO)