Zerocross Density Decomposition: A Novel Signal Decomposition Method

We developed the Zerocross Density Decomposition (ZCD) method for decomposition of nonstationary signals into subcomponents (intrinsic modes). The method is based on the histogram of zero-crosses of a signal across different scales. We discuss the main properties of ZCD and parameters of ZCD modes (statistical characteristics, principal frequencies and energy distribution) and compare it with well-known Empirical Mode Decomposition (EMD). To analyze the efficiency of decomposition we use Partial Orthogonality Index, Total Index of Orthogonality, Percentage Error in Energy, Variance Ratio, Smoothness, Ruggedness, and Variability metrics, and propose a novel metrics of distance from perfect correlation matrix. An example of modal analysis of a nonstationary signal and a comparison of decomposition of randomly generated signals using stability and noise robustness analysis is provided. Our results show that the proposed method can provide more stable results than EMD as indicated by the odd-even reliability and noise robustness analysis.