Dispersion Complex Network-Transition Entropy: A Novel Metric for Nonlinear Signal Processing

In signal acquisition, various forms of noise interference are inevitably present, and the resulting nonlinear signals severely limit the applicability of traditional signal processing methods. To address this challenge, this study proposes a novel complexity measurement metric called dispersion complex network-transition entropy (DCN-TE), which integrates the concepts of complex networks and information entropy. Specifically, we use the single cumulative distribution function values as nodes and employ Markov chains to represent the links, thereby transforming the signal into a complex network with directional weights. Then, we assess both the significance of nodes and the links to compute the DCN-TE value, and combine it with classifiers for signal processing tasks. Subsequent experiments comprehensively evaluate the performance of DCN-TE using simulated chaotic models and real hydroacoustic signals. The results indicate that compared with Lempel-Ziv complexity, permutation entropy, and dispersion entropy, DCN-TE can more rapidly and accurately capture dynamic changes in signals. Importantly, DCN-TE also exhibits optimal performance in distinguishing between different categories of chaotic models, ships, and modulation signals, thereby demonstrating its significant potential in signal processing.