Some aspects of estimating the detection rate of a periodic signal in noisy data and the time position of its components
В. И. Знак · Pattern Recognition and Image Analysis · 2009
Processing and analysis of periodic signals (including noisy ones) are considered to estimate the time position of the corresponding components including the case of frequency modulation. It is proposed to employ four steps of signal processing and analysis: weighted order statistic (WOS) filters in order to enhance the quality of both the signal itself and its components of special interest; dispersion estimation on a certain sliding basis; analysis of such an estimate as a two-dimensional black-and-white image in order to list the connected sets of nonempty clusters; and, finally, behavior analysis of cluster families in order to calculate the necessary estimates. A method is considered for the recognition of characteristic features of cluster families, and on its basis a technique is suggested for estimating the initial time for a signal to come into existence (the moment of the arrival of a wave). We give the results of processing and analysis of practical data obtained during a field trip as a part of active geophysical monitoring.