Warped signal dictionaries for the analysis of ultrasonic guided waves in long bones

In this work a new time-frequency procedure for the extraction of multimodal and dispersive guided waves (GWs) in long bones from a recorded time-waveform is presented. The proposed method is based on a basis pursuit algorithm. The ultrasonic wave is represented as a linear superposition of elementary signals, called atoms. Atoms are designed to match the spectro-temporal structures of the different propagating GWs by using a Warped Frequency Transform (WFT). The atomic distribution produces a new time-frequency representation (TFR) suitable to characterize GWs, as each atom can be associated to a particular propagating mode. An application to propagating GWs in long bones is presented to show the potential of the proposed procedure.