Auto-regressive method of detection of weak signals in noise for monitoring precursors

Abstract

Among the several premonitory indications for an impending earthquake, the observation or precursory pulses, analogous to acoustic emission before a fracture, affords a direct method of studying the seismic stability of a given area. Since the area of observation cannot be expected to be a quiet one, the technique cent change employed to detect precursory pulses embedded in a background of stationary noise must be capable of updating the information of "noise" and comparing it with that of incoming "signal". It is the aim of this paper to present a method of detection of weak signals embedded in stationary noisy background by means of array of sensors using Akaike criterion of final prediction error. Typical examples of real and artificial series are discussed to indicate the capability of the method.