Locating events using independent component analysis and Gaussian mixture models

Heather K. Hardeman-Vooys, Matt A. D. McDonald, Michael P. Lamoureux

Inspired by the work of Shamsa and Paydayesh (2019) who used Gaussian mixture models and independent component analysis to analyze microseismic data, we apply the methodology to data collected using a distributed acoustic sensor in order to detect a vehicle driving along a DAS system. We introduce Gaussian mixture models and independent component analysis. Then, we provide two examples where we calculate two independent components, the vehicle’s signal and the noise, before training a Gaussian mixture model to detect the signal. We consider two methods of training the Gaussian mixture model and compare the results. Finally, we conclude.