IFR estimation of overlapped multicomponent signals from time frequency rate distribution

Abstract

Abstract Instantaneous frequency rate (IFR) estimation from multi-component signals can be applied in a variety of applications. Viterbi algorithm (VA) on cubic phase function (CPF) is a highly performed IFR estimator for the mono-component signal. However, for the multi-component signal, when signal components are overlapped on the time frequency rate (TFR) plane, inaccurate IFRs may be tracked due to switch problem (SP) in VA and cross-terms in CPF. To overwhelm these two challenges, a directionally smoothed pseudo CPF (DSPCPF) is firstly proposed, the algorithm assumes that the IFR curve can be approximately linear along a direction within an adequately short time window, at a time instant, while the TFR points corresponding to cross-terms are dispersed in a entire TFR plane, then, TFR points of CPF along the IFR curve are integrated to enhance the auto-terms and suppress the cross-terms. In addition, to suppress the SP, the direction matrix of DSPCPF is utilized to construct an additional directional penalty function for VA. The new VA combined with DSPCPF is applied to estimate IFRs. In this way, more accurate IFRs can be obtained, and the performance of novel algorithms are validated by several artificial multi-component frequency modulated signals.