Bispectrum-based technique to remove cross-terms in quadratic systems and Wigner–Ville distribution

In the analysis of multicomponent signals using quadratic nonlinear transforms or systems, along with the desired auto-terms unwanted cross-terms are also generated. In this paper, a novel bispectrum-based technique is proposed to remove these cross-terms. The concept of accumulated modified bispectrum (AMB) is developed. The cross-term removal property of the AMB is demonstrated, and its use in quadratic system analysis in presence of additive noise is illustrated. The application of AMB in interference reduction in the analysis of multicomponent non-stationary signals using Wigner–Ville distribution (WVD) is illustrated. For performance comparison, we compute an information theoretic measure, the Reńyi entropy, which indicates that the performance of the proposed method is better than the smoothed pseudo-WVD in the analysis of multicomponent signals that have same support in time or frequency domain.