Dolphin Whistle Signal Enhancement Using Nonconvex Time–Frequency Regularized Overlapping Group Shrinkage in Warm Shallow Water

Enhancing dolphin whistle signals is a crucial preprocessing step in underwater passive acoustic monitoring (PAM), aimed at effectively improving the quality of collected signals. In warm shallow water environments, the presence of snapping shrimp noise causes typical underwater impulsive interference, significantly reducing the signal-to-noise ratio (SNR) of the received dolphin whistle signals and weakening the effectiveness of PAM systems. To address this issue, we propose a whistle enhancement algorithm based on nonconvex time–frequency (t–f) regularized overlapping group shrinkage (NCTFROGS). This algorithm leverages the group and sparse characteristics of whistle signals in the t–f spectrogram, defining the total cost function as a combination of sparse group shrinkage with t–f weighting and a nonconvex penalty function. In this framework, t–f weights are used to more effectively eliminate impulsive noise interference, while nonconvex functions promote a sparser solution in the reconstructed whistle signals. Finally, assuming that the cost function is strictly convex, we iteratively solve for the pure whistle signal using a method based on constructing auxiliary functions. Applied to whistle signals emitted by two individual Southern bottlenose dolphins, the algorithm demonstrates satisfactory enhancement effects under various types and SNRs of snapping shrimp impulsive noise interference.