Application of Common Spatial Patterns in Gravitational Waves Detection
This work addresses gravitational wave detection for astrophysics, but it is incremental as it adapts an existing method to a new domain.
The authors tackled the problem of detecting gravitational wave coalescences by applying the Common Spatial Patterns algorithm, originally used in brain-computer interfaces, to multi-detector strain data, achieving a classification accuracy of 93.72% and correctly identifying 76 out of 82 confident events.
Common Spatial Patterns (CSP) is a feature extraction algorithm widely used in Brain-Computer Interface (BCI) Systems for detecting Event-Related Potentials (ERPs) in multi-channel magneto/electroencephalography (MEG/EEG) time series data. In this article, we develop and apply a CSP algorithm to the problem of identifying whether a given epoch of multi-detector Gravitational Wave (GW) strains contains coalescenses. Paired with Signal Processing techniques and a Logistic Regression classifier, we find that our pipeline is correctly able to detect 76 out of 82 confident events from Gravitational Wave Transient Catalog, using H1 and L1 strains, with a classification score of $93.72 \pm 0.04\%$ using $10 \times 5$ cross validation. The false negative events were: GW170817-v3, GW191219 163120-v1, GW200115 042309-v2, GW200210 092254-v1, GW200220 061928-v1, and GW200322 091133-v1.