Empirical Variational Mode Decomposition Based on Binary Tree Algorithm.

Area of Science:

• Signal Processing
• Adaptive Algorithms
• Time Series Analysis

Background:

• Variational Mode Decomposition (VMD) performance is sensitive to key parameters like K, α, and τ.
• Non-stationary signals with complex components pose challenges for traditional VMD.
• Intelligent optimization algorithms for VMD parameter selection can be computationally complex.

Purpose of the Study:

• To propose an adaptive empirical variational mode decomposition (EVMD) method.
• To address VMD parameter selection issues and reduce computational complexity.
• To enhance the decomposition of non-stationary signals.

Main Methods:

• Introduced a binary tree model for adaptive decomposition.
• Utilized Signal-to-Noise Ratio (SNR) and Refined Composite Multi-scale Dispersion Entropy (RCMDE) for parameter setting (α and τ).
• Employed Least Squares Mutual Information (LSMI) and reconstruction error for cycle iteration termination.

Main Results:

• The proposed EVMD method adaptively decomposes non-stationary signals.
• Achieved reduced computational complexity (O(n^2)).
• Demonstrated good decomposition effects and strong robustness in simulations and experiments.

Conclusions:

• The EVMD method effectively solves VMD parameter selection problems.
• It offers a computationally efficient and robust alternative for analyzing complex non-stationary signals.
• The adaptive nature improves decomposition accuracy and reliability.