Radiometric Identification of Signals by Matched Whitening Transform.

Area of Science:

• Signal Processing
• Electromagnetics
• Machine Learning

Background:

• Radiometric identification aims to attribute signals to their specific sources.
• Established methods often rely on feature extraction and dimensionality reduction.
• Deep learning approaches require large feature vectors and extensive training.

Purpose of the Study:

• To develop a novel radiometric identification algorithm using whitening transformation.
• To overcome limitations of feature-dependent and computationally intensive methods.
• To enable direct analysis of raw IQ data for source attribution.

Main Methods:

• Developed a featureless radiometric identification algorithm based on whitening transformation.
• Utilized the Förstner-Moonen measure to quantify data whiteness and covariance matrix similarity.
• Employed a Majority Vote Classifier and mode function for source determination.

Main Results:

• The proposed method operates directly on raw IQ data, eliminating the need for feature vectors.
• It demonstrates superior performance compared to traditional and deep learning techniques.
• The algorithm is simpler, requires minimal training, and is non-iterative, leading to faster processing.

Conclusions:

• The whitening transformation offers an effective, efficient, and simpler approach to radiometric identification.
• This method provides a distinct alternative to maximum likelihood, Euclidean distance, and deep learning metrics.
• The featureless, direct data processing capability enhances its applicability and speed.