Gravitational Wave Mixture Separation for Future Gravitational Wave Observatories Utilizing Deep Learning

TL;DR

This paper introduces a deep learning-based method for separating overlapping gravitational wave signals in future observatories, aiming to improve detection accuracy and reduce biases in parameter estimation.

Contribution

It adapts speech separation techniques to gravitational wave data, demonstrating effective separation of overlapping signals with deep learning models.

Findings

Deep learning models successfully separate overlapping GW signals.

The method reduces biases in GW parameter estimation.

Effective for future GW observatories like the Einstein Telescope.

Abstract

Future GW observatories, such as the Einstein Telescope (ET), are expected to detect gravitational wave signals, some of which are likely to overlap with each other. This overlap may lead to misidentification as a single GW event, potentially biasing the estimated parameters of mixture GWs. In this paper, we adapt the concept of speech separation to address this issue by applying it to signal separation of overlapping GWs. We show that deep learning models can effectively separate overlapping GW signals. The proposed method may aid in eliminating biases in parameter estimation for such signals.