Description: The goal of this research project is to generate and classify infrasound propagation time series data based on natural and man-made events. Numerical methods, based on spectral methods for PDE, were used to generate long range propagation of infrasonic signals through earth’s atmosphere. Additionally, the infrasound generation process was explored using data driven approaches. Particularly, Generative Adversarial Networks (GANs) were trained to produce real valued time series with emphasis on infrasound. Trained GANs were used to generate realistic infrasound data distinguishable based on different sources and the underlying in latent space was interpolated in order to get different snapshot of same signal at different stages.

Tools: Python, Tensorflow (Keras), Google Colaboratory, Spectral methods for PDE, C++. 

Supervision: Dr A.M. Peter

This research exeprience was a summer research project in Summer 2018.