Segmented processing of electromagnetic pulse test data

A noise suppression technique that is based on the segmentation of data in both the time and frequency domains is evaluated. The segmented processing technique is based on the intrinsic behavior of electromagnetic pulse (EMP) transients. The experimentally measured data is first subdivided in the frequency domain into N_s subbands or, frequency segments, yielding N_s time-domain functions through the inverse Fourier transformation of each frequency segment. These time-domain functions may contain intervals over which noise predominates. The regions containing only noise can be eliminated by defining a subinterval of time, or time segment, for each time-domain function. The frequency-domain data can then be reconstructed through the superposition of the Fourier transforms of the time segments. The final result is a frequency-domain representation of the original signal with a lower overall noise level. The segmented processing technique is shown to significantly reduce spectral noise in typical EMP signals consisting of multiple damped sinusoids plus noise