Period Estimation of Spread Spectrum Codes Based on ResNet.

Area of Science:

• Signal Processing
• Machine Learning
• Electromagnetic Warfare

Background:

• Accurate signal identification and parameter estimation are crucial in complex electromagnetic environments for non-cooperative signal monitoring.
• Traditional methods for direct sequence spread spectrum (DSSS) signal analysis face challenges like peak energy leakage and false peak interference.

Purpose of the Study:

• To develop an efficient and accurate real-time method for detecting DSSS signals and estimating their parameters in non-cooperative scenarios.
• To address the limitations of traditional time-delay correlation algorithms.

Main Methods:

• A one-dimensional (1D) convolutional neural network (CNN) based on a residual network (ResNet) was employed for Pseudo-Noise (PN) code period estimation.
• The method treats PN code period estimation as a multi-classification problem of spread spectrum code length estimation.
• In-phase/Quadrature (I/Q) data of DSSS signals were directly input into the CNN-ResNet model for automatic feature learning.

Main Results:

• The CNN-ResNet model effectively estimated PN code lengths for DSSS signals across various signal-to-noise ratios (SNRs) from -20 to 10 dB.
• Performance was validated using Binary Phase Shift Keying (BPSK) modulated signals and tested on Quadrature Phase Shift Keying (QPSK) modulated signals.
• Analysis metrics included loss function, accuracy, recall rate, and confusion matrix, demonstrating robust generalization abilities.

Conclusions:

• The proposed 1D CNN-ResNet method accurately estimates the PN code period of non-cooperative DSSS signals.
• This approach offers a robust and effective alternative to traditional algorithms, particularly in challenging signal environments.