Overlapping Whistle-and-Click Separation for Whale Signals

To better understand whale vocalizations, it is essential to simultaneously extract the components of whistles and clicks. However, prevailing methods in whale signal processing have focused on analyzing individual whistles or clicks. In this letter, we propose a novel method for separating whistles and clicks, even when these components overlap. This method combines a time-frequency (TF)-varying Gaussian window with the synchro-compensating Chirplet transform to achieve variable anisotropy, concentrating the TF energy along the ridges of the whistle and click components in the TF representation (TFR) and computing their directional angles. By using the enhanced TFR and these directional angles, a bidirectional ridge splitter is developed to separate the components of overlapping whistles and clicks. The effectiveness of this method is validated through comparative experiments using both simulated and real whale signals.