Multiple target detection using split spectrum processing and groupdelay moving entropy

摘要

The split spectrum processing technique obtains anfrequency-diverse ensemble of narrow-band signals through a filterbanknthen recombines them nonlinearly to improve target visibility. Althoughnsplit spectrum processing is an effective method for suppressing grainnnoise in ultrasonic nondestructive testing, its application was mainlynlimited to the detection of single targets or multiple targets havingnsimilar spectral characteristics. In this paper, the group delay movingnentropy technique is introduced primarily to enhance the performance ofnsplit spectrum processing in detecting multiple targets which exhibitndifferent spectral characteristics (i.e., variations in target signalncenter frequency and bandwidth). This is likely to occur in complex,ndispersive, and nonhomogeneous media such as composites, layered, andnclad materials, etc. The analysis shows that the group delay movingnentropy method can be used effectively to select the optimal frequencynregion for split spectrum processing when detecting such targets. Basednon an iterative procedure that combines group delay moving entropy andnsplit spectrum processing, multiple targets can be identified one at antime, and subsequently eliminated by using time domain windows. Thenremoval of the dominant target improves the detection of the remainingnweaker targets. Simulation results are presented which demonstrate thenfeasibility of the multistep split spectrum processing technique forndetecting multiple targets in such materials