Statistical analysis of split spectrum processing for multipletarget detection

摘要

This work provides a statistical analysis of the performance ofnsplit spectrum processing (SSP) for the detection of multiple targetsnusing data consisting of simulated flaw signals added to experimentallynobtained backscattered grain noise. The investigation is performed underntwo conditions: known a priori target spectral characteristics (i.e.,ncenter frequency and bandwidth) which, in turn, identifies the optimalnspectral range for processing, and adaptively obtaining the processingnfrequencies using group delay moving entropy. The group delay movingnentropy method was introduced to select the optimal frequency regionsnfor SSP when detecting multiple targets. The effectiveness of thisntechnique is statistically demonstrated in this paper. The performancenis measured in terms of normalized signal-to-noise ratio (SNR) andnprobability of target detection. SSP with known target informationnyields a slightly higher probability of detection compared to SSP usingngroup delay moving entropy, while both cases achieve comparable SNRnenhancement. The SSP results were also compared with the correspondingnbandpass filter outputs, which show superior performance for SSP for anwide range of simulation parameters