Machine Learning based Glitch Veto for inspiral binary merger signals using Linear Chirp Transform

TL;DR

This paper introduces a novel method combining the linear chirp transform with convolutional neural networks to improve glitch detection in gravitational wave signals, enhancing accuracy and reducing false positives.

Contribution

It presents a new approach that integrates the chirp transform with deep learning for more effective classification of glitches versus true signals in gravitational wave data.

Findings

High accuracy in distinguishing glitches from signals

Effective use of 3D spectrograms for feature extraction

Enhanced glitch detection in LIGO and VIRGO data

Abstract

Unphysical templates for inspiral binary merger signals have emerged as an effective way to veto signals (rule out false positives) identified as glitches. These templates help reduce the parameters needed to distinguish glitches from real gravitational wave signals. Glitches are short, transient noise artifacts which often mimic the true signals, complicating gravitational wave detection. In this study, we apply the chirp transform -- a modification of the Fourier transform incorporating a linear chirp rate, denoted as $\gamma$. This method better tracks signals with varying frequencies, essential for analyzing inspiral merger events. By applying the chirp transform to gravitational wave strain time series, we generate 3D spectrograms with time,frequency and chirp axes, providing a richer dataset for analysis.These spectrograms reveal subtle features ideal for classification tasks. We leverage the use of convolutional neural networks (CNNs) to enhance the accuracy. CNNs are well suited for image based data like these, and we have optimized them to differentiate glitches from true merger signals. Our approach achieves high accuracy on both training and validation datasets, demonstrating the efficacy of combining the chirp transform with deep learning for signal classification. Ultimately we offer a refined efficient classification process by overlapping machine learning and Chirp transform, which will enhance accuracy of glitch detection from observations from detectors like LIGO, VIRGO and upcoming observatories.