Detection of uncertain seismic signals

Abstract

Current practice in seismic event detection is concentrated at the extremes of a spectrum of possibilities determined by the degree of knowledge available about the signals to be detected. At one end of the spectrum, information is available about only the frequency content of the signal. Correspondingly, simple energy detectors operating in fixed passbands are the detection method of choice: the seismological standard for detection of events over broad regions. At the other end of the spectrum, the signal is completely known. For such cases, correlation detectors (matched filters) are indicated. Correlation detectors are the emerging standard for detection of repetitive events from sources with very limited geographic extent (a few wavelengths), e.g., mines, compact earthquake swarms, and aftershock sequences. It is desirable to seek detectors that operate between the extremes, i.e., that have much of the sensitivity of correlation detectors, but more of the flexibility of energy detectors to detect signals with greater uncertainty. Subspace detectors offer one approach to achieving this trade-off. This talk describes an application of subspace detectors to detect events in a swarm with significant signal variability, and demonstrates the theoretical ability of these detectors to grade almost continuously between correlators and energy detectors.