Classification of seismic events based on short-time Fourier transform and convolutional neural network

With the increase of seismic observation data, the application of automatic processing technology in earthquake event classification, a basic work of seismic monitoring, is becoming more and more important. In this paper, 417 explosion events and 519 natural earthquake events are selected from the rich natural and non-natural seismic observation data of the Inner Mongolia Regional Seismological Network as the original data for the study. After preprocessing, such as interception and filtering, the original data is transformed into log amplitude spectrum in time-frequency domain by short-time Fourier transform, and convolution neural network with three convolution layers is used as classifier to distinguish earthquakes from explosion events. Five folds cross validation results show that the average accuracy of the algorithm used in this paper is 97.33%, and the accuracy of the test set is 98.03%. Our model has applied more original information in the classification of natural earthquake and explosion events, therefore can get a higher accuracy and better stability.