Researching the detection of continuous gravitational waves based on signal processing and ensemble learning

• Published in: Neural computing & applications Vol. 37; no. 17; pp. 10723 - 10736
• Main Authors: Pintelas, Emmanuel, Livieris, Ioannis E., Pintelas, Panagiotis
• Format: Journal Article
• Language: English
• Published: London Springer London 01.06.2025; Springer Nature B.V
• Subjects: Artificial Intelligence; Artificial neural networks; Astronomy; Black holes; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Data augmentation; Data Mining and Knowledge Discovery; Effectiveness; Ensemble learning; Gravitational waves; Image Processing and Computer Vision; Machine learning; Probability and Statistics in Computer Science; S.I.: Timely Advances of Deep Learning with applications and Data Driven Modeling; Signal processing; Special Issue on Timely Advances of Deep Learning with applications and Data Driven Modeling; Universe; Signal processing; Ensemble learning; Continuous gravitational waves detection; CNN models
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-024-10744-8

The detection of Gravitational Waves has introduced a new era for physics, astronomy, and astrophysics, unveiling new universe mysteries. Unfortunately, research studies focused on detection of Binary bursting Gravitorial Waves (B-GWs), which are produced by the rotation of binary compact objects such as black holes. Quite recently researchers attempted to detect and analyze another type of gravitational waves named Continuous Gravitational Waves (C-GWs). In contrast to the complex burst nature of B-GWs, C-GWs have an elegant and significantly simpler form while they are able to provide higher quality data for the exploration of the universe. In more detail, the direct detection of C-GWs can considerably improve our understanding of the universe by allowing researchers to examine the complexion and existence of the most extreme stars and cosmic wonders, something that was impossible up until now. However, C-GWs are much less weaker signals than B-GWs, which makes their identification a significantly challenging task. In this work, it is proposed a new framework combining signal processing techniques within ensemble learning for the effective detection of C-GWs. The key idea is to remove noise for creating a robust representation of the input data using short-time Fourier and power spectrum transformations as well as statistical tools for feeding convolutional neural network models. The individual prediction of the developed models are combined using an ensemble strategy based on a “super-learner” philosophy. In addition, the proposed framework is enforced with a data augmentation methodology for improving the prediction performance. This research work utilizes data provided by the Laser Interferometer Gravitational-Wave Observatory (LIGO). Extensive analysis confirms that the proposed framework significantly outperforms existing methods in all tested configurations, demonstrating its effectiveness.