Application of Prediction Error Filters for the Detection of Weak Teleseismic Events

This paper analyzes how the signal-to-noise ratio (SNR) for digital registrations of weak earthquakes can be enhanced by application of prediction error filters. Two prediction algorithms are considered: A linear prediction filter and an adaptive autoregressive moving average (ARMA) filter. The effects of the filter parameters on the noise suppression rate and on the quality of spectral approximation are discussed. Results are given for broad-band and short period data of the Graefenberg array. The signal-to-noise gains achieved by prediction error filtering are examined for a series of weak events.