Robust Detection and Classification of Regional Seismic Signals Using a Two Mode/Two Stage Cascaded Adaptive Arma (CAARMA) Model

摘要

The objective of this contract was to tailor TRAPS (Transient Acoustic Processing System) to the seismic event monitoring application thus providing a powerful tool for the nuclear monitoring problem. TRAPS is a three stage process. Stage One is a hybrid adaptive filter designed to perform noise elimination on each receiver component. This is beamformed and input into Stage Two which performs signal enhancement. Information resulting from Stage One/Two is used in Stage Three to calculate the location and classification of the detected event. While the original TRAPS succeeded in doubling the range of detection, study it did so with a standard beamformer. Task 1 was to ascertain if an adaptive beamformer (AB) incorporated into TRAPS would produce any gains. It was determined that TRAPS' Stage One followed by a standard beamformer was superior to an AB, yet an AB strategically placed between TRAPS' Stage One and Stage Two provided some processing gain. Keywords: Adaptive Autoregressive Moving Average (ARMA); Linear predictive residual; Pole migration; Slayed adaptive AR model; Time delay estimation maximum likelihood.