Performance tests of wavelet-based denoising methods on LIGO core-collapse supernova gravitational-wave signals

Abstract

The interpretation of gravitational-wave signals from core-collapse supernovae (CCSNe) necessitates denoising because the usual method of matched filtering is not feasible. In this work, we performed wavelet-based denoising methods on CCSN gravitational-wave signals embedded in different levels of LIGO detector noise. In particular, we used the BayesShrink and VisuShrink algorithms from the Python scikit-image package, with the Symlet, Coiflet, and Daubechies wavelet families. We compared the performance of these algorithms with that of SUREShrink algorithm reported by Lopac, et al. [Sensors 20, 6920 (2020)] and found that the BayesShrink and VisuShrink algorithms resulted in up to 47.63% lower RMSE for high signal-to-noise ratio CCSN test signals and up to 57.71% lower RMSE for low signal-to-noise ratio signals. Overall, our results suggest that wavelet-based denoising methods may provide a promising approach for GW signal denoising that are also fast-performing and readily-available.