Natural signals are often characterized by nonlinear time-frequency structures and more especially in underwater context. Underwater mammal vocalizations or dispersive phenomena are just some examples of contexts where nonlinear time-frequency structures of signal's components exist. Their analysis is of great importance for detection and classification purposes but also for phenomenon characterization. In this work, starting from the concept of warping-based time-frequency analysis, we propose a new analysis method that combines the properties of the waping transform with the concept of compressive sensing. It provides a more accurate characterization of nonlinear time-frequency structures in terms of the estimation of their parameters. Results provided for simulated data prove the interst of this new approach with respect to the spectrogram-based method.