CONSTRUCTION OF A BAND-VARIABLE FILTER FOR EARTHQUAKE RECORDS PROCESSING USING ANALYTICAL MODE DECOMPOSITION METHOD

摘要

Earthquake signals are typical non-stationary signals with time varying spectrums.In order to obtain its time-frequency features,an improved analytical mode decomposition(AMD)method is proposed to process earthquake ground motion records.Firstly,AMD computational formula and its numerical implementation are given,which can decompose a signal into components with mutually exclusive frequency contents.This AMD algorithm is then improved so as to decompose signals with time-varying bisecting frequency.A band-variable filter is constructed by recursively using the improved algorithm with proper bisecting frequencies.Continuous wavelet transform is used to obtain time-frequency spectrograms of earthquake signals and suggest time-varying bisecting frequencies.Finally,the band-variable filter is used to extract components containing intensive energy from earthquake signals.Analysis results of synthetic signals and earthquake recordings show that the proposed filter is efficient and accurate,and helpful to understand time-frequency domain energy distribution of non-stationary signals including,but not limited to,earthquake signals.