Choice of Computer-Generated Random Numbers Whose Statistics Match Those of Experimentally Measured Ambient Noise.
Abstract
The family of fluctuation-sensitive processors known as the WlSPR family, has been demonstrated to significantly improve the detection of submerged sources in a variety of experimental cases. However, since the experiments do not cover all possible ranges of signal and environmental parameters, it is necessary to perform simulations in order to develop a more complete understanding of processor performance under different conditions. The choice of computer-generated random numbers (RNs) whose statistics match those of experimentally measured ambient noise for such simulations is discussed in the present report. Both the statistics of voltage time series and the statistics of power decibel levels of specific frequency bins calculated by fast Fourier transform have been used for comparison between the computer-generated RNs and experimentally measured ambient noise. Experimental ambient noise data from the Black Sea and the Pacific Ocean have been used. The results show that the statistics of computer-generated uniform RNs and Gaussian RNs match those of experimental ambient noise data very well. Therefore, the RNs should provide realistic simulation of ambient noise in simulations of WlSPR processor performance.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 04, 1996
- Accession Number
- ADA319187
Entities
People
- Jacob A George
Organizations
- United States Naval Research Laboratory