LIGO Core-Collapse Supernova Detection using Convolution Neural Networks

TL;DR

This paper presents a CNN-based method combining time-frequency analysis for detecting gravitational waves from core-collapse supernovae, achieving high accuracy especially at higher SNRs.

Contribution

It introduces a novel CNN approach trained on simulated signals and noise, utilizing spectrograms from STFT and Q-transform to improve CCSNe GW detection.

Findings

Near 100% detection rate for SNR > 0.5

STFT outperforms Q-transform at low SNRs

Effective CNN model for CCSNe GW detection

Abstract

Core-Collapse Supernovae (CCSNe) remain a critical focus in the search for gravitational waves (GWs) in modern astronomy. Their detection and subsequent analysis will enhance our understanding of the explosion mechanisms in massive stars. This paper investigates a combination of time-frequency analysis tools with convolutional neural network (CNN) to enhance the detection of GWs originating from CCSNe. The CNN was trained on simulated CCSNe signals and Advanced LIGO (aLIGO) noise in two instances, using spectrograms computed from two time-frequency transformations: the short-time Fourier transform (STFT) and the Q-transform. The algorithm detects CCSNe signals based on their time-frequency spectrograms. Our CNN model achieves a near 100% true positive rate for CCSNe GW events with a signal-to-noise ratio (SNR) greater than 0.5 in our test set. We also found that the STFT outperforms the Q-transform for SNRs below 0.5.