Recognition of Noisy Radar Emitter Signals Using a One-Dimensional Deep Residual Shrinkage Network

Area of Science:

• Electrical Engineering
• Computer Science
• Signal Processing

Background:

• Traditional radar emitter signal recognition methods struggle with accuracy in noisy environments due to obscured signal features.
• Developing robust methods for feature learning from noisy signals is crucial for reliable radar applications.

Purpose of the Study:

• To propose a new radar emitter signal recognition method using a one-dimensional deep residual shrinkage network (1D DRSN).
• To improve the accuracy of radar emitter signal recognition, especially in the presence of significant noise.
• To enable automatic feature learning from raw 1D signal data without requiring signal processing expertise.

Main Methods:

• A one-dimensional deep residual shrinkage network (1D DRSN) was developed for radar emitter signal recognition.
• The method incorporates a soft thresholding function to eliminate unimportant features, with thresholds adaptively set by an attention mechanism.
• The 1D DRSN learns signal features directly from 1D data, bypassing the need for manual feature engineering or dimension conversion.

Main Results:

• The 1D DRSN demonstrated high recognition rates for radar emitter signals even in noisy conditions.
• Experimental validation confirmed the effectiveness of the 1D DRSN across various noise types.
• Comparative analysis showed superior performance of the 1D DRSN over other deep learning methods.

Conclusions:

• The proposed 1D DRSN offers a powerful and automated approach for radar emitter signal recognition in challenging, noisy environments.
• The network's ability to learn features directly from 1D data and eliminate redundant information contributes to its high accuracy and robustness.
• The soft thresholding function and attention mechanism are key components enabling the network's effective performance.