Instantaneous Frequency Estimation of Multicomponent Signals Characterized by Interfering Linear Trajectories

摘要

The paper deals with signals that include nonstationary components. For these signals, the estimation of the instantaneous frequency of each component is especially difficult in time intervals where its trajectory interferes with another one, for example by crossing it. The attention is mainly paid to linear trajectories, seeing that any trajectory can be generally approximated as linear in the nearby of the crossing point. The authors propose a method based on the use of a two dimensional subspace of the chirplet transform, known as bowtie subspace, that allows accurate estimates of the instantaneous frequency even in the presence of crossing trajectories. The use of a restricted subspace allows a reduced computational burden and less time consumption with respect to several approaches based on advanced multiparametric transforms, typically exploited when multicomponent nonstationary signals are analyzed. The performance granted by the proposed method is assessed through a number of tests performed on simulated, emulated, and real signals.