Reconstruction of gravitational wave signals in flexible bases

The gravitational wave signals for events detected by LIGO and Virgo can be reconstructed in two complementary ways: template-based analysis using physical waveform models, or model-agnostic methods independent of source physics assumptions. BayesWave analysis implements the latter and reconstructs signals as a sum of frame functions using either a wavelet or a chirplet basis. Wavelets are particularly effective in reconstructing signals from high-mass binary black hole (BBH) systems, where the time-frequency volume of the signal is relatively compact. Conversely, chirplets - which are generalized wavelets that allow for linear frequency evolution - perform better in recovering signals with lower signal-to-noise ratios (SNR) and larger time-frequency volumes. However, the optimal conditions for using either chirplets or wavelets - particularly in terms of total mass and SNR - remain to be systematically explored. Events reconstructed using BayesWave in the LIGO-Virgo-KAGRA Gravitational Wave Transient Catalogs have so far exclusively used the wavelet basis for waveform consistency tests, either excluding or poorly reconstructing events with low SNR or high time-frequency volumes. This talk presents an injection study using simulated BBH signals to compare BayesWave reconstruction faithfulness between wavelet and chirplet bases. Our study aims to establish quantitative guidelines for selecting the optimal frame function in BayesWave based on total mass and SNR, enabling improved signal reconstruction across a broader range of gravitational wave events.