Multi-chaotic signal identification employing a causal cross-correlation neural network

Area of Science:

• Complex Systems Analysis
• Nonlinear Dynamics
• Signal Processing

Background:

• Chaos identification is vital for understanding complex systems.
• Existing methods struggle with temporal causality, noise robustness, and identifying multiple chaotic signals.

Purpose of the Study:

• To develop a robust method for identifying multi-chaotic signals amidst noise.
• To overcome limitations of current chaos identification techniques.

Main Methods:

• Examined 15 chaos models, enhancing the Swish-modulo map with acquisition-enhanced Bayesian optimization.
• Developed a causal cross-correlation network featuring learnable wavelet transform, time-frequency feature fusion, and spectral-inversion-free fast Fourier transform-based cross-correlation (SFFT-CC).

Main Results:

• Achieved 96.67% chaos identification accuracy under 20 dB noise conditions.
• The SFFT-CC method demonstrated superior accuracy (up to 5.42% improvement) over dilated convolution.
• Realized a 43-fold reduction in training time compared to traditional methods.

Conclusions:

• The proposed causal cross-correlation network effectively identifies multi-chaotic signals with high accuracy and efficiency.
• The SFFT-CC approach offers a significant advancement in noise-robust chaos identification for complex systems.