Skeleton-based reassignment of nonstationary signals spectrogram

Many practical applications require ridge curves detection and enhancement as they allow for instantaneous frequency estimation of nonstationary signals. Unfortunately, the reconstruction of those curves becomes a difficult task whenever multicomponent signals having non-separable modes have to be analyzed. To address this issue, this paper introduces a novel binary model for the spectrogram of multicomponent signals, and applies skeleton to enhance IF curves. To further improve the skeleton-based reassignment, a convolution-deconvolution scheme, that is directly derived from a specific evolution law for signal time-frequency (TF) distribution, is also defined. Experimental results show that the skeleton-based approach can reproduce IF curves of separable modes accurately. In addition, it allows to outperform classical amplitude-based reassignment method in the case of non separable modes.