An EMD with Masked WVDs for Cross-terms Reduction in Non-stationary Signals

Signals commonly found in real-world applications are non-linear, multi-component, and non-stationary. For simplicity, they are analyzed and processed in time-frequency (TF) domain. Wigner-Ville distribution (WVD) is a method that provides time-frequency representation (TFR) of a signal with high-sharpness resolutions. However, its bilinear nature causes false frequency components called “cross-terms” to occur between the real components of the multi-component signal. WVDs of mono-component but non-linear signals also suffer from cross-terms problem. Empirical mode decomposition (EMD) is a data-driven algorithm used to decompose a multi-component signal into subcomponents called intrinsic mode functions (IMFs). This paper applies WVD masking for each extracted signal components obtained from EMD to ensure that the cross-terms will be suppressed, especially if there are non-linear IMFs. Also, masked WVDs using different linear TF methods are also investigated to obtain the comparative results. The results revealed that by using the masked WVDs to EMD, the cross-terms were satisfactory removed.