Autoregressive Search of Gravitational Waves: Denoising

Abstract

Because of the small strain amplitudes of gravitational-wave (GW) signals, unveiling them in the presence of detector/environmental noise is challenging. For visualizing the signals and extracting its waveform for a comparison with theoretical prediction, a frequency-domain whitening process is commonly adopted for filtering the data. In this work, we propose an alternative template-free framework based on autoregressive modeling for denoising the GW data and extracting the waveform. We have tested our framework on extracting the injected signals from the simulated data as well as a series of known compact binary coalescence (CBC) events from the LIGO data. Comparing with the conventional whitening procedure, our methodology generally yields improved cross-correlation and reduced root mean square errors with respect to the signal model.