Scholarlink

<p>Future gravitational wave (GW) observatories, such as the Einstein Telescope, are anticipated to encounter overlapping GW signals, presenting considerable obstacles to GW data processing techniques, including signal identification and parameter estimation. In this letter, we propose a scheme of combining deep learning and Bayesian analysis to disentangle overlapping GW signals. The deep learning part takes a data-driven approach that employs an encoder-separation-decoder framework which is powerful enough to extract the shape of the signal even when the GW signals completely align. The Bayesian analysis part takes the matched filtering technique to extract the amplitude of the GW signals. Our scheme can facilitate the utilization of existing GW detection and parameter estimation methods for future instances of overlapping strain. This methodology effectively reduces biases in parameter estimation when handling multiple intertwined signals. Remarkably, this marks the first known instance where deep learning has been successfully utilized to disentangle overlapping GW signals.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Combine deep learning and Bayesian analysis to separate overlapping gravitational wave signals

  • 期刊论文

  • 发表日期：

    2025年3月14日

查看全文
  • Cunliang Ma,
  • Weiguang Zhou,
  • Zhoujian Cao,
  • Mingzhen Jia

摘要

Future gravitational wave (GW) observatories, such as the Einstein Telescope, are anticipated to encounter overlapping GW signals, presenting considerable obstacles to GW data processing techniques, including signal identification and parameter estimation. In this letter, we propose a scheme of combining deep learning and Bayesian analysis to disentangle overlapping GW signals. The deep learning part takes a data-driven approach that employs an encoder-separation-decoder framework which is powerful enough to extract the shape of the signal even when the GW signals completely align. The Bayesian analysis part takes the matched filtering technique to extract the amplitude of the GW signals. Our scheme can facilitate the utilization of existing GW detection and parameter estimation methods for future instances of overlapping strain. This methodology effectively reduces biases in parameter estimation when handling multiple intertwined signals. Remarkably, this marks the first known instance where deep learning has been successfully utilized to disentangle overlapping GW signals.