engleski

Combining Time-Frequency Signal Analysis and Machine Learning with an Example in Gravitational-Wave Detection

This paper presents a method for classifying noisy, non-stationary signals in the time-frequency domain using artificial intelligence. The preprocessed time-series signals are transformed into time-frequency representations (TFRs) from Cohen’s class resulting in the TFR images, which are used as input to the machine learning algorithms. We have used three state-of-the-art deep-learning 2D convolutional neural network (CNN) architectures (ResNet-101, Xception, and EfficientNet). The method was demonstrated on the challenging task of detecting gravitational-wave (GW) signals in intensive real-life, non-stationary, non-Gaussian, and non-white noise. The results show excellent classification performance of the proposed approach in terms of classification accuracy, area under the receiver operating characteristic curve (ROC AUC), recall, precision, F1 score, and area under the precision-recall curve (PR AUC). The novel method outperforms the baseline machine learning model trained on the time-series data in terms of all considered metrics. The study indicates that the proposed technique can also be extended to various other applications dealing with non-stationary data in intensive noise.

non-stationary signals ; time-frequency representations ; artificial intelligence ; machine learning ; convolutional neural networks ; gravitational waves

nije evidentirano