Detecting and suppressing dolphin clicks in whistle signals using adaptive matched filtering and interpolation algorithm.

Abstract

This paper describes a method for removing click disturbances from dolphin whistle signals. The algorithm is based on click end point detection, disturbance removal, and missing whistle segment interpolation, and enables both click suppression and whistle recovery from within mixed signals. Considering the impact of dolphin whistles on click detection, a Gabor filter is used to preprocess the spectrogram of the mixed signals. An adaptive matched filtering algorithm based on the Teager-Kaiser energy operator is then applied to detect the click end points, allowing the clicks to be removed from the mixed signals. Finally, a least squares interpolation algorithm based on a linear prediction model is introduced to recover the missing whistle segments, thereby achieving click suppression. Simulations are conducted to explore how the whistle intensity influences the performance of the click detection method. The robustness of the whistle interpolation algorithm is then tested using real data from three cetacean species. The results show that the proposed click detection algorithm achieves higher accuracy and greater robustness than traditional methods, and the click suppression scheme performs better than other compared denoising schemes under different signal-to-interference ratios.