Swarm decomposition: A novel signal analysis using swarm intelligence

A novel approach in non-stationary signal decomposition, namely swarm decomposition (SWD), that fosters rules of biological swarms to address non-stationary signal analysis, is presented here. Cornerstone of SWD is the swarm filtering (SwF), a processing approach envisioned by a swarm-prey hunting. Under proper parameterization, the output of iterative applications of SwF results in an individual component of the input signal. To control the method, the relationships between "hunting" parameters and particular responses of SwF are extracted using a genetic algorithm. SWD consists of successive applications of iterative SwF under different "hunting" parameters, so as the existing components to be extracted. The SWD is evaluated through its application to non-stationary multi-component (both synthetic and real-life) signal decomposition. The results obtained by SWD are compared with the respective ones obtained by empirical mode decomposition, wavelet-based multiresolution analysis and an iterative approach based on eigenvalue decomposition of the Hankel matrix, achieving higher accuracy in correctly isolating the components of the analyzed signals in the most cases. The promising results pave the way for a new approach in signal decomposition with a wide range of application potentialities. (C) 2016 Elsevier B.V. All rights reserved.