Rapid joint detection and classification with wavelet bases via Bayes theorem

Abstract

The discrete wavelet transform (DWT) is currently being used for seismic-event detection and classification in the New England region. The DWT forms a new basis set for picking out, from a data stream, important features of a seismic event: time, energy, and predominant period of the first, peak, and last waveforms. Classification of these events from their features into one of the following classes, teleseisms, regional earthquakes, near earthquakes, quarry blasts, and false triggers, is accomplished with conditional class densities derived from training data. This algorithm is tested for detection and classification performance on the New England Seismic Network (NESN) of Weston Observatory of Boston College. This detection algorithm exhibits a likelihood of detection two times greater than STA/LTA under typical wideband network constraints in arbitrary conditions at NESN stations. Classification of seismic events via this method achieves an approximately 70% correct identification rate relative to a human viewer over a broad range of data test sets.